ESSAYS ON LAND PROPERTY RIGHTS IN CAMBODIA: EMPIRICAL ANALYSIS
A DISSERTATION SUBMITTED TO THE GRADUATE DIVISION OF THE UNIVERSITY OF HAWAI‘I AT MĀNOA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
IN
ECONOMICS
AUGUST 2012
By
Kouland Thin
Dissertation Committee: Sumner La Croix Sang Hyop Lee Nori Tarui Timothy Halliday Alice G. Dewey
Keywords: Land property rights
ACKNOWLEDGEMENTS
I am thankful to the East West Center and the Pacific Center of Excellence in the
Elimination of Disparities for supporting me financially during the course of my studies at the University of Hawai‘i at Mānoa (UHM). I also would like to thank professor
Sumner La Croix, who is my dissertation adviser, for his direct involvement in this dissertation and his support throughout my academic program. During this dissertation writing, I have benefited greatly from my dissertation committees. I would like to thank professor Sang Hyop Lee, professor Timothy Halliday, professor Nori Tarui, and professor Alice Dewey for their comments and advice on this dissertation.
I would like to extend my gratitude to George Cooper and Jaynie Estermann for their support during my dissertation writing. I also wish to thank my friends at Hale
Mānoa and various individuals at UHM community who support me spiritually during my stay in Hawai‘i. I am thankful to my parents and my parents-in-law who always provide me good food and encourage me to fight for this degree. The love and support from my husband and my son have made a substantial contribution to the completion of this dissertation.
ii
ABSTRACT
This dissertation consists of four chapters on land rights in Cambodia. The first chapter focuses on historical background of land rights in Cambodia. De jure private property rights in land were first established in 1863. The civil wars of the 1970s led to an abolition of land property rights and it was again reintroduced in 1989 in which households could register their plots through sporadic land registration. The combination of a poorly functioning land registry system and rising demand for arable land led to a sharp increase in conflicts over land rights in the 1990s. To address these problems, the
Cambodian government carried out a series of land policy reforms in the 1990s and introduced a systematic land registration in 2002 to speed up land registration in the country. The second chapter empirically estimates the effect of land property rights on investment in land by farm households in Cambodia using the 2003-2004 Cambodia
Socio-Economic Survey (CSES). In this chapter, I use matching regression techniques, in particular Propensity Score Matching (PSM), to account for selection effects in household participation in land titling program. The empirical analysis indicates a positive link between de jure land rights and land investment. The third chapter estimates the effect of de jure land rights on household consumption. Instrumental variable techniques are used to account for endogeneity of de jure land rights. The results show that the choice to obtain de jure land rights results in a statistically significant increase in household consumption. The fourth chapter explores the relationship between de jure land rights and child health in households of land owners in Cambodia. The 2003-2004
CSES which is used to estimate this effect contains information of 8,745 children ages
iii
below 6 years old. The findings show that children of titled parcels have higher height- for-age z-scores than children of untitled parcels.
iv
TABLE OF CONTENTS
Acknowledgments………………………………………………………………………...ii
Abstract…………………………………………………………………………………...iii
List of tables………………………………………………………………………….....viii
List of figures……………………………………………………………………………...x
Summary of chapters……………………………………………………………………...1
Chapter One: Land rights in Cambodia…………………………………..……………….5
1.1 Introduction………………………………………………….………………...5
1.2 The French colonial period (1863-1953)……………………………………...7
1.3 Post colonial period (1953-1975)…………………………………...………..10
1.4 Land property rights during the Khmer Rouge regime (1975-1979)……..….11
1.5 Land property rights during the PRK (1979-1989)…………...……………...12
1.6 Land property rights from 1989 till the present………………………..…….14
1.7 Sustainability of the cadastral database……………………………….……..21
1.8 Conclusion…………………………………………………………….……..23
1.9 Appendix 1.A Process of getting land titles…………………...…………..…29
1.10 References…………………………………………………………………..31
Chapter Two: Land rights and investment in land……………………………….....……33
2.1 Introduction…………………………………………………………..………33
2.2 Literature review………………………………………………………..……35
2.2.1 Capacity of land governance institutions…………………….…….38
2.2.2 Cultural diversity and context-specific conditions………………...39
v
2.2.3 Issues in the empirical studies………………………..……….……40
2.3 Prospects for enhancing land tenure security in Cambodia………………….42
2.3.1 Agriculture………………………………………………………....42
2.3.2 Land investment in Cambodia (1979 to 2004)………..…………...44
2.4 A model of tenure security and investment …………………………………46
2.4.1 Case 1: Land tenure security is exogenous………………………...48
2.4.2 Case 2: Land tenure is endogenous…………………...………...... 49
2.5 Empirical analysis………………………………………………………...... 50
2.5.1 Data………………………………………………………………...50
2.5.2 Farm characteristics………………………………………………..51
2.5.3 Household characteristics………………………………………….53
2.5.4 Empirical models and econometric issues……………………...... 53
2.5.5 Propensity score matching (PSM)……………………...………….55
2.5.6 Estimation of propensity scores………………………………..…..56
2.5.7 Estimation of the effect of land rights through PSM……………....59
2.6 Conclusion………………………………………………………………...... 64
2.7 Panel A: Land title…………………………………………………………...71
2.8 Panel B: Land title and application of receipt………………………………..74
2.9 Panel C: Application of receipt…………………………………………...... 77
2.10 References…………………………………………………………………..80
Chapter Three: Property rights in land and household consumption………………….....85
3.1 Introduction…………………………………………………………………..85
3.2 Land reform and poverty reduction………………………………………….87
vi
3.3 Land reform and poverty reduction in Cambodia………………………...... 90
3.4 Empirical analysis and econometric issues………………..………………....93
3.4.1 Empirical model…………..………………………………...... 93
3.4.2 Regression estimates…...……………….………………………….97
3.4.3 Data…………...……………………………………………..……..98
3.4.4 Household consumption…...……………….………………………98
3.4.5 OLS regression results…...…………………………………….....100
3.4.6 Robustness of the OLS regression results………………………...104
3.5 Conclusion……………………………………………………………….....110
3.6 References………………………………………………………………...... 119
Chapter Four: The impact of property rights in land on child health………...………...124
4.1 Introduction……………………………………………………...………….124
4.2 The impact of property rights on children’s health…………………………126
4.3 Children’s health in Cambodia…………………………...………………...129
4.4 Empirical analysis…………………………………...……………………...130
4.4.1 Data……...………………………………………………………..130
4.4.2 Estimation strategy…...….……………..…………………………135
4.4.3 Results………...……………………………...…………………...137
4.4.4 Weight-for-height……...…………………..……………………..139
4.4.5 Height-for-age……...……………………...……………………...140
4.5 Conclusion……………………..…………………………………………...143
4.6 References…………………………...……………………………………...154
vii
LIST OF TABLES
Table Page
1.1 Cambodia: Population and real GDP………………….………………………...….26
1.2 Average price of residential land in urban areas……………………..……………..27
1.3 Average price of agricultural land in rural areas…………………..………………..27
1.4 Legal status of agricultural land in 2004 and 2009…………………………..……..28
2.1 Land investment: 1979-2004…………………………..……………………………66
2.2 Farm characteristics: Summary statistics………...………..………………………..68
2.3 Household loans: Summary statistics…………………………….…………………69
2.4 Household characteristics: Summary statistics………………….……...…………..70
2.5a Propensity score estimation result for titled farm regression…..…....…………...…71
2.5b Result from PSM: Land titles…………...………………...………………………..72
2.5c Treated vs. controls (land titles)…………………………...………………………..73
2.6a Propensity score estimation result of a combined measure regression...………...…74
2.6b Result from PSM: Combined measure………………………….………………….75
2.6c Treated vs. controls (combined measure)……………………….………………….76
2.7a Propensity score estimation result of application receipt farm regression.…...... ….77
2.7b Results from PSM: Application receipts …………………………...………..……..78
2.7c Treated vs. controls (application receipts)……………………...…………………..79
3.1 Village characteristics: Summary statistics………………………..………….……112
3.2 Household loans: Summary statistics………………………………………..…….113
3.3 Household characteristics: Summary statistics……………………………………114
viii
3.4 Land titles and household consumption: OLS and instrumental variable regressions....115
3.5 A combined land rights and household consumption: OLS and instrumental variable regressions……………………………………………………………………………....116
3.6 Application receipts and household consumption: OLS and instrumental variable regression…………………………………………...……………………….………….117
3.7 Regression of land rights on all instruments and exogenous variables………….....118
4.1 Descriptive statistics of z-scores for child nutritional status in Cambodia, 2003-
2004……………………………………………………………………………………..144
4.2a Characteristics of children and their families by type of document..……………...146
4.2b Characteristics of children and their families: Summary statistics..………………147
4.3a Land title: Weight-for-height…………………...……………...………………….148
4.3b Combined land rights measure: Weight-for-height…………...…………………..149
4.3c Application receipt: Weight-for-height……………………………………………150
4.4a Land title: Height-for-age…………………………...…………………………….151
4.4b Combined land rights measure: Height-for-age…………………..………..……..152
4.4c Application receipt: Height-for-age ………………...……………………………15
ix
LIST OF FIGURES
Figure page
1.1 Procedures for sporadic land registration…………………………...………..……...29
1.2 Procedures for systematic land registration…………………………...... ……..….30
2.1 Land investment: 1979-2004…………………………………………..………….…67
2.2 Estimated propensity scores: Land titles…………………………………………..…72
2.3 Estimated propensity scores: Combined measure…………………………………....75
2.4 Estimated propensity scores: Application receipts…………………………………..78
4.1 Distributions of weight-for-height and height-for-age z-scores in Cambodia, 2003-
2004………………………………………………………………………………..……145
x
Summary of Chapters
Chapter One: Land rights in Cambodia
This chapter describes the historical background of land rights in Cambodia from the French colonial period until the present. Cambodia has changed its political regime many times since gaining independence from the French in 1953. There have been at least five different property rights regimes counting from the colonial period until the present, leading to contemporary confusion over land rights. The modern concept of land holding did not exist until the arrival of the French in 1863. A system of land registration was established by the French and continued after the country gained independence from
France in 1953. The civil wars of the 1960s and 1970s and the emergence of the communist Khmer Rouge as the governing authority for certain parts of the country from the late 1960s until the late 1990s and for the entire country from 1975 to 1979 generally involved a collectivization of land and abolition of individual property rights. The
People’s Republic of Kampuchea (PRK), which succeeded the Khmer Rouge following
Vietnam’s invasion of Cambodia in 1978, retained collectivization, while reintroducing some limited private rights to land but initially not land ownership. This relaxed form of collectivization slowly gave way in the 1980s to de facto private property rights, which became formalized in a 1989 reform in which households could register possession for agricultural land and ownership for their residential plots through sporadic land registration. The combination of a poorly functioning land registry system and rising demand for arable land in an expanding economy led to a sharp increase in conflicts over
1
land rights in the 1990s. In an attempt to address these problems, the Cambodian
government carried out a series of land policy reforms in the 1990s, culminating in the
adoption of a new land law in 2001 and the implementation of systematic land
registration (SLR) operated under a western-donor-assisted land management and administration project (LMAP) starting in 2002 and still continuing. The government’s
long-term goal is to register all Cambodian lands, and to enforce recording of their
transfer on its official registry in order to maintain records of land holders for tax
collection purposes and to facilitate land market transactions.
Chapter Two: Land rights and investment in land
This chapter empirically estimates the effect of land rights on land investment by
farm households in both urban and rural areas in Cambodia after the reform in 1989. The
mix of farm households having different land rights raises the possibility of estimating the effects of land reform on investments made by farm households in both titled and un- titled farm lands. The 2003-2004 Cambodia Socio-Economic Survey (CSES) includes information on household land investment and the legal status of a household’s farmlands, thereby facilitating analysis of this relationship. In this chapter, I use matching regression techniques, in particular Propensity Score Matching (PSM), to account for selection effects in household participation in Cambodia’s program to convert de facto
land claims into formal de jure land rights supported by a national land registry. The
empirical analysis indicates a positive link between de jure land rights and land
investment. Investment on titled farms is higher than on untitled farms. The findings of
2
this study provide additional support for there being a link between formal land rights and
positive agricultural outcomes in low-income environments.
Chapter Three: Property rights in land and household consumption
This Chapter empirically estimates the impact of de jure land rights on household
well-being, which is measured by household consumption. Previous microeconomic
studies have posited that de jure land rights reform in developing countries had the potential to reduce poverty and enhance economic growth. De jure enforcement and titling of arable lands are often posited as factors that facilitate low-income households climbing out of poverty by allowing land holdings to serve as collateral for loans to improve farm productivity and to smooth consumption expenditures of farm households.
This Chapter utilizes data from the 2003-2004 CSES to explore the link between de jure land rights and household consumption. The finding from this study provides new evidence on the impact of formalizing land rights in low-income countries, a topic which has not been widely studied yet. The Chapter’s empirical analysis uses three different measures of de jure land rights: the households that possess land plots with land titles, land titles or application receipts, and just application receipts. Instrumental variable techniques are used to account for endogeneity of de jure land rights. The empirical estimations show that the choice to obtain de jure land rights results in a statistically significant increase in household consumption in regression estimates that do not account for endogeneity of de jure rights. The impact becomes larger when the regression includes instruments for de jure land rights.
3
Chapter Four: The impact of property rights in land on child health
This chapter explores the relationship between de jure land rights and child health in households of land owners in urban and rural areas in Cambodia. There are currently few studies that focus on how the establishment of land rights affects children’s health.
Galiani and Schargrodsky (2003, 2004) argued that land titles could have positive indirect impacts on child health because they lead to more investment in housing and household structures. Vogl (2007) asserted that the positive effect of land rights in urban areas of Peru on child nutrition might come from increases in labor force participation, housing investment, and credit access. This study maintains that the effect of land rights on child health could take place via other channels, such as the increase in household consumption, land investment, and housing investment. This study estimates this relationship for a sample of households drawn from the 2003-2004 Cambodia Socio-
Economic Survey (CSES) that includes information on 8,745 children ages 0-5. Weight- for-height and height-for-age are used as measures for children’s health. Three different measures of de jure land rights are utilized in this analysis: the household that possesses land plot with land title, land title or application receipt, and just an application receipt.
The analysis uses instrumental variable regression techniques to account for the selection effects in a household’s decision to participate in a land titling program to convert de facto land claims into formal de jure land rights. The results from this analysis show that children of titled parcels have higher height-for-age z-scores than children of untitled parcels but there are no statistically significant differences in weight-for-height z-scores outcomes in a regression with instrumentation of land rights.
4
CHAPTER ONE
LAND RIGHTS IN CAMBODIA
1.1 Introduction
Cambodia has changed its political regime many times since gaining
independence from the French in 1953. There have been at least five different property
rights regimes counting from the colonial period until the present, leading to
contemporary confusion over land rights. The modern concept of land holding did not exist until the arrival of the French in 1863. With colonization came a modern bureaucracy that established and administered a system of land registration allowing an individual to claim property rights to a land parcel, a right that was formerly the exclusive domain of the king (So, 2009; Hartman, 2006; Chan & Sarthi, 2002; Sik, 2000).
A system of land registration was established by the French and continued after the country gained independence from France in 1953. The civil wars of the 1960s and
1970s and the emergence of the communist Khmer Rouge as the governing authority for certain parts of the country from the late 1960s until the late 1990s and for the entire country from 1975 to 1979 generally involved a collectivization of land and abolition of individual property rights. The People’s Republic of Kampuchea (PRK), which succeeded the Khmer Rouge following Vietnam’s invasion of Cambodia, retained collectivization, while reintroducing some private rights to land. Households were allowed to farm on either collective land or individually held plots within the collective
(So, 2009; Chan, Tep, & Sarthi, 2001; Sik, 2000). This relaxed form of collectivization slowly gave way in the 1980s to de facto private property rights, which became
5
formalized in a 1989 reform in which households could register their plots through
sporadic land registration. The transition from collective land holding to household
claims on land through occupation (de facto rights) to de jure property rights in 1989
with a weak cadastral managerial system led to problems of tenure security for
uneducated and poor households who were not integrated into the new land registration
system (So, 2009). Registration of land ownership was costly and inaccessible to most
cultivators with de facto rights. As a result, only a small proportion of households who
applied for land ownership title in 1989 and the 1990s actually received land titles. The
majority of landholders initiated the land registration process by filing an application but
did not follow through with land registration. They obtained only application receipts from cadastral officers (So, 2009; Chan & Sarthi, 2002).
The combination of general lawlessness, a poorly functioning land registry system
and rising demand for arable land in an expanding economy led to a sharp increase in
conflicts over land rights in the 1990s. In an attempt to address these problems, the
Cambodian government carried out a series of land policy reforms, culminating in the
adoption of a new land law in 2001 and the implementation of systematic land
registration (SLR) operated under a land management and administration project (LMAP)
starting in 2002 (World Bank, 2002). Under a 2001 land law, a land holder can obtain
title through either a revised sporadic land registration (known as non-mandatory
registration) or systematic land registration (SLR, also known as mandatory land
registration). The government’s long-term goal is to register all Cambodian lands, and to enforce their transfer through its registry in order to maintain records of land holders for tax collection purposes and to facilitate land market transactions (So, 2009).
6
1.2 The French colonial period (1863-1953)
Prior to the French colonial period, all land belonged to the king. The king
allowed his people to cultivate his land. Practically, each plot of land belonged to the
person who tilled it. Since the population was still small, people could move freely from
one place to another and assume de facto ownership of unoccupied land (Chan & Sarthi,
2002; Sik, 2000; Greve, 1993).1 Hartman (2006) concludes that pre-1863 land rights were essentially usufruct rights where land holders had rights to use land and pass it to heirs as long as they continued to keep it in production. With low population densities and growth, there was no need to regulate land rights. Thus, there were no cadastral documents that recorded information on land ownership. Rights were established by local acceptance rather than formal systems. Governing authorities also did not rely upon precise specification of land rights as taxes were imposed on harvests rather than on the quantity or value of the land (So, 2009; Thion, 1993).
De jure private property rights to land were first introduced in 1863 when the country became a French colony. The French government imposed a formal system of private property rights in order to tighten its control over the country, to collect taxes, to advance and protect French business interests, and to sell unoccupied lands for large- scale plantation projects (So, 2009; Thion, 1993). Lands were categorized as “crown land, public land, reserved land, and alienable land”. Ordinary people were only entitled to hold alienable land.2 Vacant lands could be settled but settlers were restricted to claim
what they could use (Cooper, 2002, p.15).
1 Chandler (1993) estimated that the population of Cambodia was less than one million people in the middle of the nineteenth century. 2 Alienability was not previously a feature of Cambodian land rights. 7
To proceed with land registration, the government created land governance bodies
to administer land registration and to oversee the production and maintenance of a land
book, i.e., a land database. The Department of Cadastre, which was later changed to
Department of Cadastre and Topography in 1908, was established in 1896. The government also began to form a commune committee in 1908 to facilitate the process of creating a land book. In establishing the land book, the commune chief took an important role in the process vis-à-vis the people in his commune by certifying the land owner and identifying the location of the land (Lim, 1997).3
After the implementation of the land registration, there were few changes between
1884 and 1920. The land registration system did not work particularly well during this time due to resistance to registration by the elite and village peasants and to weak land governance institutions. The Cambodian elite simply disagreed with this new land law.
The elimination of slavery and establishment of individual land property rights outraged the elites whose interests depended on traditional arrangements of labor exploitation and taxation of harvests. They also felt insecure because their positions were changed from exploiting the people to being paid civil servants of the French colonial regime. Further, the new land laws changed the basis of taxation from the harvest to the quantity of their land, and village peasants realized few if any public benefits from their tax payments.
The amount of tax paid was estimated to be around 30 percent of the value of the peasant’s annual crop; French administrators made little investment in infrastructure or social development with the revenues (So, 2009; Chandler, 1993).
Despite some resistance, the implementation of land registration continued with the adoption of the civil code in 1920 and the Fixed Asset Declaration (FAD) in 1925.
3 Lim Voan was the director of the national land title department during the 1990s. 8
The civil code of 1920 recognized both possession and ownership land rights. A land holder could file for a possession land right through fixed asset registration. The commune authority was obligated to send the fixed asset registration to the district land office and leave a copy for landholders. This land holders’ copy of fixed asset registration was recognized as a possession right. The formal ownership title could be issued to a land holder only after the FAD was registered in the land book at the district land office. The registration of FAD in the land book required a land survey and a legalization process involving identification of the land owner (So, 2009; Lim, 1997).4
There were different procedures in transferring land with possession rights and ownership rights. The transfer of land with possession rights occurred at the commune office, whereas the transfer of land with ownership rights required a written property transfer form and was registered at the district land office (So, 2009). The possession status given under the 1920 civil code was equivalent to ownership in terms of rights that except the possession status was forfeited when the land was left unused for a number of years. This did not include the case where lands were left fallow (Cooper, 2002).
By 1930 most rice growing plots were registered. The majority of the land was split into plots of smaller than 5 hectares, while large land holdings associated with plantation projects had also been created. Meanwhile, many people continued to hold their lands according to the customary land holding system. People who sold their lands moved into the forest (the common land) and established new land plots (Sik, 2000;
Greve, 1993). This customary land holding was also recognized by the 1920 civil code. A
4 The sub-national structure of the Cambodian government runs from village, commune, district, and to province level. 9 land holder who peacefully occupied unregistered land for five consecutive years became eligible to apply for ownership rights (So, 2009).
1.3 Post colonial period (1953-1975)
The legacy of a mixture of customary and formal private property rights in land was continued by the Cambodian governments who assumed power after independence in 1953. Despite initial resistance, the elites began to accept the system of titled private property as they perceived that it was a form of investment and a store of wealth (So,
2009; Williams, 1999). Land registration continued, with land holders able to pursue a claim to either possession rights or ownership rights. At the same time, claiming land ownership rights through occupation also continued. The coexistence of modern and traditional land holdings created land conflicts when the government was unable to protect vulnerable groups who had not titled their lands. Powerful elites and military officials seized the opportunity to reap benefits from undocumented land holders by issuing themselves land ownership rights on lands held by peasants under customary rights. These actions culminated in a peasants’ revolt in 1967. The revolt was thereafter suppressed by government forces. Many people who were involved in this revolt sought refuge in the forests, where the communist rebels also hid (So, 2009; Thion, 1993).
The period from 1953 to 1970 is marked as the beginning of increasing land inequality and land concentration in Cambodia. Starting from 1953, wealthy households slowly accumulated more fertile rice fields (So, 2009). The statistics from 1962 indicated that 53.74 percent of peasants held land smaller than 5 hectares, 24.82 percent held land between 5 and 10 hectares, and 21.45 percent held land larger than 10 hectares (Lim,
10
1997). By 1970, the number of land-owning peasants had shrunk as the number of
landlessness increased to about 20 percent. The growing inequality was partly due to
political instability and a weak land governing institution in which people with political
connections and more resources seized this opportunity and took control of fertile land
(So, 2009; Kiernan & Boua, 1982).
Economic hardship together with administrative corruption and the failure of the
state to protect village peasants from powerful persons created an unfavorable situation
for the government led by Prince Sihanouk. This situation led to the collapse of his
government in 1970. The republican government led by General Lon Nol from 1970 to
1975 still acknowledged private land property rights but new land registration was
impeded by internal political chaos. In the early 1970s, Cambodia became a war zone
contested by armed rebels, the Khmer Rouge Force (KRF), and the American-sponsored
republican government (So, 2009). Land registration was not completed when the KRF
took control in 1975. Only 10 percent of the land was registered in the land book,
whereas the rest of the land was held under FAD (Lim, 1997).5
1.4 Land property rights during the Khmer Rouge regime (1975-1979)
Private property rights were abandoned when the KRF took full control of the country in 1975. The new regime, named “Democratic Kampouchea”, was led by Pol Pot who had a vision, influenced by Mao’s ideas, to build a collectivized, agriculturally- based economy. The Khmer Rouge leader wanted to use agriculture as a means to build the country. To achieve his utopian vision, an all-encompassing land collectivization
5 There are no precise data on the number of plots registered each year from the start of land registration in 1863 to its end in 1975. The figure provided in the text is a rough estimate by the officer who worked for the cadastral office before the full Khmer Rouge takeover in 1975. 11
system was implemented. The KRF evacuated hundreds of thousands of people from
every city and town to carry out back-breaking labor in the countryside without providing
adequate food to sustain people’s lives and energy. People were forced to work on rural
public works projects, such as building dams or irrigation systems to achieve the
government’s goal of increasing rice productivity (So, 2009; Twining, 1989). As part of
its collectivization of agriculture, the KRF eliminated virtually all forms of property
rights and supporting institutions. They destroyed all cadastral records, including
cadastral maps and land ownership titles (Sik, 2000).
The KRF could not achieve its vision due to poor management, lack of inputs to
nourish cultivated land, and no motivation from the Cambodian people to work hard in
the new economic environment. Consequently, dams and irrigation systems were poorly
built and the goal to greatly increase agricultural productivity was never accomplished.
The failure of this endeavor led Khmer Rouge cadres to search for and purge the
“enemies of the revolution” (So, 2009; Becker, 1998). Over a 44-month period, approximately 1.7 million Cambodian people perished from disease, starvation, and execution (So, 2009).
1.5 Land property rights during the People’s Republic of Kampuchea (1979-1989)
The Khmer Rouge regime was toppled by Vietnam’s invasion of Cambodia in
1978. A new government, the People’s Republic of Kampuchea (PRK), was immediately
established. It was led by former KR, controlled by Vietnam and was communist in
character. This government initially depended heavily on support from Vietnamese soldiers and aid from the eastern communist bloc to suppress the remaining KRF and
12
non-communist forces that did not recognize the legitimacy of the PRK government (So,
2009; Frings, 1993).
In early 1979 after the new government took power, most of country’s rice fields
were left fallow because there were not enough inputs and agricultural equipment to farm
them and much of their infrastructure had been destroyed or needed a great deal of repair.
In late 1979, PRK officials together with their Vietnamese advisors decided to retain the
collectivization system in a relaxed form. The government expected that the
collectivization system would allow the state to equally distribute resources and to feed its cadres and soldiers in the long run. The government further hoped that getting people to work in the fields would allow rice production to resume, help resolve the country’s food crisis, and save the economy (Gottesman, 2003).
From 1979 to 1989, all land belonged to the state. Households were allowed a small piece of land for residential purposes, but could not obtain any legal documents documenting their possession and rights. The rights to residential land during the PRK regime were essentially de facto land rights in which a household established use rights through occupation (Sik, 2000; Chan & Sarthi, 2002). Cultivated land was collectively
managed by a solidarity group. A plot of land could be farmed collectively or privately
within the collective. Each solidarity group was composed of 10 to 15 families (So,
2009).6 Solidarity groups were classified into three different types: (1) land was collectively held by the group and cultivated by group members; output was shared by group members; (2) land was held by the group but the group was divided into a sub- group of 3 to 5 families; each subgroup managed the production process in which
6 The number of families in a solidarity group varied according to the population in each region. 13
produce was divided by subgroup member; and (3) the group’s land was divided into
plots that were allocated to an individual family and privately farmed (Lim, 1997).
During the 1980s, the collectivization system slowly failed for a number of
reasons. First, the government lacked human resources to support and promote the
collective system. Second, people did not have motivation to work because output was distributed to some group members who did not work. Further, materials to support cultivation were not distributed equally. Seeds, fertilizers, fuels, and other agricultural equipment were mostly made available only to model solidarity groups. Third, for fear of a political backlash the government did not take strict actions against group members who were able to work but did not participate. The experience from the extreme collectivization system imposed during the Khmer Rouge regime made people suspicious of the PRK’s leadership motives, and they objected to actions intended to restore the system (So, 2009; Frings, 1993). Slowly, on a de facto basis, a private property system began to creep back in during the 1980s, in which people “sold” or exchanged their land to others with simple unwritten verbal agreements or using simple documents that were authenticated by village or commune chiefs as was done in transferring possession rights in pre-Khmer Rouge years (So, 2009).
1.6 Land property rights from 1989 till the present
In 1989, the government implemented a land reform program that established de jure private rights to residential and arable land. Politics and rural conditions were primary reasons behind the program. The PRK government faced political challenges
following the withdrawal of Vietnamese troops in 1989 and a subsequent reduction of aid
14
from the Soviet Bloc, as it had to confront the remaining Khmer Rouge forces with fewer
resources. To deal with the Khmer Rouge armed resistance, the PRK had to broaden
conscription, a measure with the potential to provoke popular resentment. The land
privatization program was intended to compensate the people for the unfavorable military
draft policy. At the same time, the government realized that land collectivization could
not produce enough output to feed the population. The lack of private land property rights
discouraged investment in existing farms and clearance of new land for cultivation.
Although agriculture production increased in the late 1980s, the government projected that Cambodian famers could not produce enough to meet the needs of Cambodia’s then- rapidly increasing population (So, 2009). Thus, the policy of moving toward private property rights was adopted both to stabilize the country’s politics and to increase agricultural productivity to meet the demand for food by the growing population.
The 1989 land privatization program had its legal roots in constitutional amendments, in sub-decree number 25 on providing house ownership to the Cambodian population, and instruction number 3 on implementation of land use and management policy, all adopted in 1989.7 A new land law was adopted in 1992 (So, 2009).8 Land
ownership rights established before 1979 were not recognized by any of these measures.
Instruction number 3 granted transferable ownership rights for residential land which did not exceed 2000 square meters and possession rights for cultivated land which did not exceed 5 hectares. A farmer could claim a possession right to cultivated land as long as the farmer had used the land one year before the date of declaration of the
7 Council of Ministers, Instruction No. 03 SNN on Implementation of land use and management policy, 03 June 1989. Council of Ministers, Sub-Decree ANK 25 on Providing house ownership to the Cambodian population, 22 April 1989. 8 National Assembly, Land Law, 10 August 1992. 15
instruction. In addition, returnees from abroad and refugee camps were eligible for
residential and cultivated land according to the availability of land. The rest of the
collective land was redistributed to households based on family size and availability of
land (Sik, 2000).
Following the land privatization in 1989, issues related to land distribution and land registration arose. According to the instruction, land was reallocated according to the
size of the family and availability of land within the village. Application of this rule
would result in variations in land holdings across villages due to differences in land
availability and population, but families of a given size within a particular village were
supposed to receive equal allocations. In practice, other factors influenced official
allocations of land, as authorities kept good farmland for themselves, their relatives, and
friends. Moreover, the sporadic land registration—a voluntary land registration system
whereby land holders could register their holding and receive a title—began operations in
1989 and was widely viewed as ineffective and a mechanism for validating the large land
holdings that had been acquired by powerful groups and individuals (So, 2009).
The main issue behind the ineffectiveness of the sporadic land registration
program was the unwillingness of the government to register land without officials being
paid fees and bribes so large that no one but the upper middle class and the wealthy could
afford registration. Also the cadastral department was poorly equipped and did not have
enough staff to handle the influx of applications for land titles (So, 2009). Another important issue was the sporadic land registration program’s complicated procedures. An
application for land title must complete a number of steps before land title can be issued;
some can be completed at the communal level, while others require action by officials at
16
the district level (Chan et al., 2001). Figure 1.1 provides a flow diagram of the titling
process. Each stage of an application process requires an approval from the cadastral
officer in charge. This multi-step bureaucratic procedure delays the application process as
it provides numerous opportunities for the responsible officer to freeze an application and
ask for a bribe to proceed (So, 2009; Frings, 1993).
During the first two years in which households could initiate the titling process
(1989-1990), millions of Cambodian started the process of registering their land holdings.
In 1990 local governments issued receipts for applications covering 3.7 million plots; applications tailed off in 1991, as the total number of receipts issued grew to over 4 million (So, 2009).9 Due to the reasons articulated above, only a small percentage of
households completed the registration process and received a title for their lands.
Through 1997, less than 15 percent of households starting the registration process had
received a land title. Most of the titled plots were in or near the capital city of Phnom
Penh. Other households with application receipts began to view receipts as a weak form
of land title, as the receipt was an official document that noted their claim while not
providing specifications of its extent or boundaries and not being part of an official
registry (Chan et al., 2001; Sik, 2000).
The rapid expansion of the Cambodian economy in the 1990s along with high
rates of population growth put enormous pressure on both official and unofficial systems
of land holdings. Economic growth was on average 7 percent annually in the years 1994-
2004 (WB, 2006). Cambodia’s population had also been surging, increasing from about 6
9 The application receipt is a little piece of paper given by cadastral officers when a household first applies for a land titling certificate or possession certificate. The application is first processed at the commune chief office. The receipt contains a permanent address of the applicant, the registration number of the receipt, date of application, name of land holder, size of land, category of land, and a signature from a cadastral officer. 17
million people in the early 1980s to more than 11 million people in 1998 (Chan & Sarthi,
2002). Table 1.1 illustrates that population growth in the 1990s averaged 2.9 percent.
Together these two factors pushed up the price of residential and farm land in Cambodia.
The price of residential land in urban areas rose from $10/m2 in 1995 to more than
$20/m2 in 2000. The price of cultivated land also rose sharply, from $0.07/m2 in 1995 to
$0.19/m2 in 2000 (Chan & Sarthi, 2002). Tables 1.2 and 1.3 provide lists of prices of
residential and cultivated land in both rural and urban areas. The price of land varies
according to the location and size of land.10
The rise in land values led land speculators to buy up lands in both rural and urban areas. It also induced, in combination with ineffective land registries and lands held only by initial application receipts or without documents, an increased number of land- related conflicts in the juridical and other institutions. Between 1992 and 1993, the
United Nations Transitional Authority in Cambodia (UNTAC) received many complaints
related to boundary issues and land grabs by civilians and soldiers (So, 2009;
Ledgerwood, 1998). The appeals court received 1,325 land dispute cases between 1995 to
mid-1999; in 2001 alone, the court received 1,310 cases (Cooper, 2002). The 2003-2004
Cambodia Socio-Economic Survey (CSES) revealed that approximately 2 percent of land
plots were engaged in litigation and a 2005 USAID report indicated that slightly more than 1 percent of the population was engaged in land conflicts (So, 2009).
Further, various documents indicate that land holdings became more concentrated after the 1989-1992 land reforms. A small survey in three villages conducted in 1996 showed that the top 10 percent of households occupied 34 percent of village land, whereas the bottom 40 percent held just 9 percent (So, 2009; CDRI, 1997). A report from
10 The price of land in the tourist regions is mostly high in comparison to other areas. 18
Oxfam GB in 2007 from a survey of 433 villages revealed even greater concentration: 12
percent of households owned 72 percent of village lands while 67 percent of households
accounted for less than 8 percent (So, 2009).
The Cambodian government adopted land policies in 2001 designed to both
lessen the frequency of land disputes and reduce land concentration. In its “Statement on
Land Policy”, the government set out its policy on land management, administration, and
distribution. It called for the creation of a land inventory, enhancing the capacity to resolve land disputes, registering all land under SLR or under a revision of the 1989-2001
system of sporadic land registration, and strengthening land tenure security. The SLR—
mandatory land registration—was to be accomplished under the auspices of the Land
Management and Administration Project (LMAP), an endeavor that drew technical
support from various donors and was primarily funded by a loan from the World Bank.
The project’s goal was to complete the land registration within its set 15-year life span,
2002-2017 (WB, 2002).
The legal basis for the SLR was the land law adopted in 2001 by the national assembly of Cambodia and various sub-decrees. The 2001 land law was written to remedy some flaws in the 1992 land law. It requires the utilization of a “cadastral index map” for all plots in order to clarify their boundaries. Private ownership is allowed for both residential and cultivated land. It allows people who already occupied land peacefully before the adoption of the law to complete a five-year possession term in order to claim ownership rights to their land. Following the effective date of the 2001 land law new occupation of a plot of land by clearing it or settling down on it with this leading to a possession right and then ownership was prohibited. The law does, however, have
19
provisions for social land concessions in which the government allocates empty state land
to landless families for building a house or cultivation (So, 2009).
The SLR is an important part of the current land policy. The program is designed
to register all land parcels in Cambodia, to put all land ownership under government’s
supervision through creation of cadastral index maps, and to maintain cadastral land
records. Figure 1.2 sets out the various steps by which a land holder can title holdings
under the SLR. SLR is a mandatory land registration program in which all plots in a
selected location are registered together; normally an SLR location is a commune and
most/all lands in that commune are registered. The teams from the cadastral office, which
are composed of registration specialists and land surveyors, are sent to the selected
location to gather information about land owners, to perform land surveys, and register land into the cadastral system (So, 2009).
In addition to SLR, land can still be registered through a revised sporadic land
registration. A sub-decree embodying a revised version of the program was issued in May
2002.11 Some of the procedures listed in Figure 1.1 were removed. Step two in Figure
1.1, the formation of a subcommittee, was taken out. If there is a land conflict or a
counter claim, then the application process is halted at the district cadastral office until
the issue has been settled. If no problems occur, then an application receipt is issued to
the applicant and the date of land survey and demarcation is set. When the land survey
and demarcation are completed, the district cadastral officer signs a form documenting
the land holder’s identity and the land’s boundary. This form is then disclosed to the
public for 30 days. The land is then registered by the district cadastral staff into a
sporadic land registration map, and the application together with all land documents are
11 Sub-decree 48 ANK on sporadic land registration, 31 May 2002. 20
forwarded to the provincial or city cadastral office where information on the land claim is
recorded into the land registry. Finally, the application is sent to the General Department
of Cadastre and Geography (GDCG) for issuance of title.12 This revised version of the
titling procedures is not much different from the previous one.
According to the Decision issued in 2006, the time to issue a land title under SLR
is longer than under sporadic land registration. It takes 5 to 7 months for title to be issued
under SLR and 2 to 3 months under sporadic land registration. The fee for issuance of a
land title under SLR in rural areas was set at $0.25/1000 m2 for agricultural land, whereas
the fee for issuance of a land title under sporadic land titling registration in all provinces
and municipalities was approximately $2.5.13 The fee charged under sporadic land
registration is not clearly defined. It does not clearly distinguish the different amounts
charged on land of various sizes. This flaw makes it difficult to actually calculate the fee paid for a title issued under sporadic land registration.
During its first phase (2002 to 2007), the LMAP operated in 11 of Cambodia’s 24
provinces, surveyed more than one million land plots, and issued titles for more than
800,000 land plots (So, 2009). Table 1.4 provides information on agricultural land plots
in Cambodia with registered titles in 2003-2004 and 2009.14 It indicates that about 21
percent of arable land was held with titles in 2004, but only about 11 percent in 2009.
The remainder was either held with various types of documents or no documents. Some
of the differences in the two surveys may be due to confusion between receipt and title,
as some households who reported in the 2004 survey may have treated their receipts as
12 Ibid. 13 Decision on cadastral service standards, 21 December 2006. 14 The data in Table 1.4 are from CSES 2003-2004 and 2009.
21 titles. The 2009 survey may be more accurate because households were asked to show their documents during the survey interview.
1.7 Sustainability of the cadastral database
Since 2002, the World Bank and other donors to Cambodia have devoted significant amount of resources to LMAP to register land systematically and develop land rights institutions. The goal of this effort is to increase tenure security, which can have a positive impact on overall well-being and poverty alleviation in the country. Owners, whose names are recorded in the land registry, receive de jure rights that the government has promised to enforce. However, the majority of the people (a smaller percentage in urban areas and a larger percentage in rural areas) do not register a transfer when transfer of land ownership. This results in the central land registry becoming outdated and negatively impacts the effort to develop an effective system of de jure land rights in
Cambodia (So, 2009).
There are several issues that make households opt out of formal land transfers.
First, complicated procedures and a high transfer fee discourage households from registering land transfers. A formal land transfer starts from the district cadastral office which reports a land transfer to the Ministry of Land Management, Urban Planning and
Construction (MLMUPC). After the transfer form is completed and signed by the district chief, a tax amounting to 4 percent of the sale’s value must be paid at the provincial tax department. The provincial department of the MLMUPC must verify the transaction agreement and forward the transfer form together with the tax receipt to GDCG for final approval of the transaction (Chan & Sarthi, 2002). Several bureaucratic stages must be
22
completed to finalize the transfer and it is impossible to move from one stage to another
without paying unofficial fees (So, 2009). Moreover, households also revealed that high
transfer taxes, which account for 4 percent of the sale’s value, discourage them from
making official land transfers. A formal land transfer eats up almost 30 percent of land
value whereas the informal transfer is much cheaper and less time consuming (So, 2009).
The second issue stems from a weak judicial system enforcing de jure rights.
People have no trust in the judicial system providing them justice when a land dispute
erupts (Chan & Sarthi, 2002). Many poor people turn instead to local officials or politicians to solve the issue (So, 2009). The third issue is related to a routine practice which is widely accepted by the people and government officers. People often go to the commune office to conduct their transfer and believe that the written agreement with a signature and stamp from a commune chief is enough to certify their transaction. This is the most common form of land transfer implemented and a commune chief does not regularly report land transactions to the district office (Chan & Sarthi, 2002). This form of land transfer was also implemented in the period before the Khmer Rouge regime in which the land registered under FAD was transferred at the commune office.
Due to the reasons articulated above, it is very difficult to shift to a formal registration system since government protection of land tenure is weak. This attenuates incentives to participate in the formal registration system. At the same time, the informal system also provides people with some forms of tenure security. Some people reason that land registration is unnecessary because the price of land is low and no one would file a claim on their land as they continue to utilize it (Chan & Sarthi, 2002). The coexistence of the two systems greatly impacts current government efforts to build a sustainable
23
cadastral system since titled land gets transferred through the informal system as untitled
land. The outdated cadastral system can be a source of land conflicts and can severely
impact land tax collections and land transaction market (So, 2009).
1.8 Conclusion
The coexistence of de facto and de jure land holding systems in Cambodia created land tenure insecurity for households who were not integrated into the de jure land holding system (especially poor and less-educated persons) in the 1960s and again, in the
1990s and 2000s. In both instances, the national government started a land titling program that was not widely utilized, at least in part due to weak government institutions registering land and enforcing de jure rights. In both cases, land grabs occurred. After gaining independence from the French in 1953, the government continued the land registration program that was imposed by the French in 1863, but failed to take appropriate measure to protect undocumented land holders. Land grabs by government and military officers who registered ownership claims on undocumented land culminated in a peasants’ revolt in 1967. On the eve of the 1970s political turmoil, the number of peasants who owned land had shrunk as landlessness increased to about 20 percent. The issue reemerged in 1989 when the PRK government reintroduced private property rights in land after their abolition during the Khmer Rouge regime. The combination of economic expansion and rapid population growth in the 1990s increased land tenure insecurity for peasants without registered titles.
The incumbent government set out its policy on land management, administration, and distribution to lessen land conflicts and land concentration in 2001. The
24 government’s goal was to register all land plots in the country under SLR or sporadic land registration. Land registration increased during this period, but the system has worked poorly because few owners of titled land have registered transfers of their titles through the formal system. In other words, the LMAP was successful in building an initial registry of land holding but not in updating the registry as the land has been transferred. This negates the ultimate goal of cadastral sustainability. Most households holding titled lands opt out of the cadastral office’s formal system of transfers due to bureaucratic procedures and corruption inside the cadastral office. In the absence of a working land registry, an existing norm/ practice has been continued: Households go to the commune office to transfer lands. Without additional efforts by the government to enforce land transfer through cadastral offices, the central database becomes outdated and fails to serve the functions for which it was established.
25
26
Note: Residential and agricultural land price. Adapted from “Land transactions in Cambodia: An analysis of transfer and transaction records,” by Chan and Sarthi, 2002, CDRI, p. 28-31.
27
Table 1.4
Legal status of agricultural land in 2004 and 2009
Type of documents 2004 CSES 2009 CSES Land tile certificate 43.53% 22.17% Application receipt 49.60% 23.45% Land investigation paper 2.85% 0.37% Paper from local authority 0% 37.37% Rental contract 0.82% 0.80% Other 3.20% 15.84% Note: The CSES 2004 indicates that about 52.65 percent of cultivated land was held with various types of documents, whereas CSES 2009 shows that 50.93 percent of cultivated lands was held with various types of documents. Source: All data are from CSES (2003-2004) and (2009).
28
Appendix 1.A
Figure 1.1
Procedures for sporadic land registration
a) An applicant has to make application to his/her commune chief in a standard form. Completed application forms are certified by the commune chief and forwarded to Office of Land Management, Urban Planning, Construction and Land (OLMUPCL) at the district. The owner receives a receipt for his/her application.
b) A sub-committee is formed consisting of the commune chief as its chair, the village chief (of the applicant’s village), one or two officials of the staff of the OLMUPCL (district cadastral office) and one older person from the community whose knowledge and authority in the village is respected.
c) The sub-committee, in consultation with the applicant, fixes a date for site inspection. Owners of neighboring lands to the applicant’s land parcel are also called to ensure correct and mutually acceptable boundary demarcation during the measurement process.
d) Prior to the visit, an announcement is made by the district chief about the intent of the applicant to title his/her land. A notice period of 10 to 15 days is allowed for any party to raise objections or contest the claim. A notice to this effect is displayed on or near the site (for example, in the village concerned, or at the relevant commune office or district office).
e) During the visit, the applicant must erect signposts at appropriate places in order to demarcate his/her land and also identify the shape of the land. Neighbors can raise objections, if they have any, at this stage. The land survey is completed upon satisfactory completion of the boundary demarcation. A standard form is then filled in and duly authenticated by neighbors and others on the sub-committee. The background of the applicant, his/her family details and the mode of land acquisition (for example, through inheritance or purchase), are noted. Any plantations or construction on the land are also recorded. In the event that one or more of the neighbors is a state authority, their representative(s) are called for the boundary inspection.
f) In the event of a dispute concerning the applicant’s claim, the processing of the claim is halted until the dispute is resolved. Otherwise, the form is signed by the commune chief in his/her capacity of chairperson, and forwarded to the district OLMUPCL Cadastral Office with a recommendation to issue a certificate. The relevant documents are then forwarded to the district chief.
g) In the case of agricultural land, the district chief can approve issue of the certificate if s/he is satisfied. S/he can call for clearance from other concerned parties if s/he deems it necessary, for example, from forestry or fisheries departments. In the case of residential land, the chief of the district cadastral office recommends the matter to the district chief for verification, who in turn refers the case to Department of Land Management, Urban Planning, Construction and Cadastre (DLMUPCC) at the provincial level. This department then submits the papers to the provincial
governor for final approval for issuing a certificate.
Note: Procedures for land registration. Adapted from “Land tenure in Cambodia: A data update ,” by Chan et al., 2001, CDRI, p. 27-38 .
29
Figure 1.2
Procedures for sporadic land registration
Note: Procedures for land registration. Adapted from “Land tenure in Cambodia: A data update,” by Chan et al., 2001, CDRI, p. 27-38.
30
References
Becker, E. (1998). When the war was over: Cambodia and the Khmer Rouge revolution. New York: Public Affairs.
Cambodian Development Resource Institute (CDRI). (1997). Rethinking food security: Report on the preliminary findings of the CDRI/UNRISD study on food security and coping strategies in Cambodia. Phnom Penh: CDRI
Chandler, D. (1993). A history of Cambodia. Sydney: West View Press.
Chan, S., Tep, S., & Sarthi, A. (2001). Land tenure in Cambodia: A data update (CDRI Working Paper 19). Phnom Penh: CDRI.
Chan, S., & Sarthi, A. (2002). Land transactions in Cambodia: An analysis of transfers and transaction records (CDRI Working Paper 22). Phnom Penh: CDRI.
Cooper, G. (2002). Land policy and conflict in Cambodia. Unpublished manuscript.
Frings, V. (1993). The failure of agricultural collectivization in the people republic of Kampuchea 1979-1989. Clayton, Australia: Center for Southeast Asian Studies, Monash University.
Gottestman, E. (2003). Cambodia after the Khmer Rouge: Inside the politics of national building. New Haven: Yale University Press.
Greve, H. S. (1993). Land tenure and property rights in Cambodia (paper prepared for UNTAC). Phnom Penh.
Hartman, N. (2006). Cambodia land law: Its impact on current trends in landlessness and land inequality. Mentis Vita, VI-ii, 115-125.
Kiernan, B., & Boua, C. (1982). Peasants and politics in Kampuchea 1942-1981. London: Zed Press.
Ledgerwood, J. (1998). Rural development in Cambodia: The view from the village. In F. G. Brown, & D. G Timberman (Eds), Cambodia and the international community: The quest for peace, development, and democracy (pp. 127-147). New York: Asia Society.
Lim, V. (1997). Land regime in Cambodia. Unpublished manuscript.
National Institute of Statistics. (2003-2004). CSES 2003-2005 data management manual. Phnom Penh: Cambodia
31
Sik, B. (2000). Land ownership, sales, and concentration in Cambodia: A preliminary review of secondary data and primary data from four recent surveys (CDRIWorkingPaper16). Phnom Penh: CDRI.
So, S. B. (2009). Political economy of land registration in Cambodia. (Doctoral dissertation). Retrieved from ProQuest Dissertations and Theses. (Accession Order No. AAT 3359039)
Thion, S. (1993). Watching Cambodia: Ten paths to enter the Cambodian tangle. Bangkok: White Lotus.
Twining, C. H. (1989). The Economy. In K. D. Jackson (Ed.), Cambodia 1975-1978: Rendezvous with death (pp. 109-150). Princeton: Princeton University Press.
United States Agency for International Development (USAID). (2005). Cambodian village directly affected by land conflict and located in conflict affected area. Phnom Penh: Cambodia.
Williams, S. (1999). Land rights and access in Cambodia. Phnom Penh: Oxfam GB.
World Bank. (2002). LMAP project appraisal documents report no. 22869-KH. Phnom Penh: World Bank.
______. (2006). Cambodia halving poverty by 2015? Poverty assessment 2006. Phnom Penh: World Bank.
32
CHAPTER TWO
LAND RIGHTS AND INVESTMENT IN LAND
2.1 Introduction
The acceleration of economic growth and an increase in population density in the
1990s augmented land tenure insecurity in Cambodia, especially for poor and less
educated groups. The 1989 land reform and the subsequent introduction of a sporadic
land titling registration program have, however, been unable to resolve land tenure
insecurity in the country. Registration of land ownership proved to be costly and inaccessible to most cultivators with de facto rights. Consequently, only a small percentage of people applying for ownership or formal possession of their land received formal possession or ownership documents (So, 2009; World Bank, 2002; Chan, Tep, &
Sarthi, 2001). The combination of a poorly functioning land registry system and rising demand for arable and commercial land in an expanding economy induced sharp increases in conflicts over land rights. The rise of conflicts affected the lives of many
Cambodians and became a major issue that threatened people’s livelihoods and raised the possibility of pushing already impoverished farmers deeper into poverty (So, 2009).
In an attempt to address these problems, a range of land policy reforms were carried out in the 1990s, culminating in the adoption of a new land law in 2001 and the implementation of systematic land registration (SLR) under a Land Management and
Administration Project (LMAP) starting in 2002. This land reform effort was aimed at improving security of tenure, a component of land rights that reformers expected would lead to increased access to credit markets, incentives to make larger investments in the
33
land, and to improvements in agricultural productivity. A World Bank report (2002)
argued that poor and vulnerable groups would gain the most from the reforms, and that
the reforms would reduce poverty.
Empirical studies of the effect of de jure land rights on land investment have not
always supported the effects articulated in economic theories of de jure property rights.
Abdulai, Owusu, and Goetz (2011), Goldstein and Udry (2008), Do and Lyer (2006),
Jacoby, Li, and Rozelle (2002), Alston, Libecap, and Mueller (1999), and Besley, (1995)
found a positive link between land rights and land investment, while Jacoby and Minten
(2007), Braselle, Gaspart, and Platteau (2002), Place and Migot-Adholla (1998), Place and Hazell (1993), Migot-Adholla, Hazell, Blarel, and Place (1991) did not. Other studies maintained that their inconclusive results were due to the small sample sizes utilized in most published empirical studies (Deininger & Jin, 2006) and the failure of most studies
to adequately account for endogeneity of land rights (Brasselle et al., 2002). Small
sample sizes reduce the precision of estimates and tend to bias estimated coefficients
towards zero (Deininger & Jin, 2006; Deaton, 1997). Endogeneity of household choices
to register de facto claims and obtain de jure land titles is accepted as a critical issue in
this literature, as households who have made larger capital investments in their lands
have a larger incentive to complete the bureaucratic process for obtaining a title from the
government land registry to protect their earlier investments.
Cambodia has the potential to be a good case study of the impacts of de jure land
rights on land investment by farm households because its land titling program has been
operated long enough to make an assessment and, more importantly, the program has
been implemented during a time of political stability and economic growth. This long
34 political stability facilitated program implementation unlike an earlier land titling program that was interrupted by civil war. The smooth implementation of a land titling program is very important for an agrarian economy like Cambodia where the majority of the rural population depends heavily on cultivation to survive. Most importantly, household survey data that incorporate information on household land titles are available for Cambodia. The 2003-2004 Cambodia Socio-Economic Survey (CSES) includes information on household land investment and the legal status of a household’s farmlands, thereby facilitating analysis of this relationship. In this chapter, I use matching regression techniques, in particular Propensity Score Matching (PSM), to account for selection effects in household participation in Cambodia’s program to convert de facto land claims into formal de jure land rights supported by a national land registry. This
PSM method is very useful when matching treated and untreated observations involving many covariates, as it yields a weighting scheme that provides an unbiased estimate of the treatment effect (Dehejia & Wahba, 2002).
The empirical analysis in this study provides a result supporting the theory of land rights: Investment on titled farms is higher than untitled farms. The findings of this study also provide support for the link between land rights and agricultural outcomes in low- income environments amidst the inconclusive results on this relationship offered by earlier studies.
2.2 Literature review
Secure and transferable land rights have been documented in the economics literature as a key to economic growth (Field, 2007; Deininger & Jin, 2006; Carter &
35
Olinto, 2003; De Soto, 2000; Jacoby et al., 2002; Besley, 1995; Binswanger, Deininger,
& Feder, 1995; Pinckney & Kimuyu, 1994). Three linkages between land rights and
economic growth stand out. First, secure land rights enhance investment incentives. This
is a traditional view that highlights the gains resulting from freedom from expropriation
(Besley, 1995; Alchian & Demsetz, 1973; Demsetz, 1967). Households are more willing
to undertake investment if they expect to reap benefits from their investments (Besley,
1995; Brasselle et al., 2002). Second, a clear proof of legal ownership of the land rights
allows legitimate land owners to access lower interest rates in credit markets, thereby
raising returns on investment projects, as lenders become confident that their investment
is secured (Feder & Nishio, 1998). Third, clear land rights facilitate secured transfers of
land from inefficient to efficient users. They allow land to be owned by those who value
it the most and also assure buyers that their lands will not be confiscated in the future
(Jacoby & Minten, 2007).
Securing land rights through registration of land titles and enhancing rights to
transfer and their social benefits have been widely discussed in the past two decades
because empirical findings on this linkage have been inconclusive. Studies from most
Asian countries provide evidence supporting the theory of a positive relationship between
land property rights and agricultural outcomes. A prominent study of Thailand indicated
that land titles have positive and significant effects on land improvements and access to
credit markets (Feder, Onchan, Chanlamwong, & Hongladarom, 1988). In Vietnam,
improvement in land rights, such as rights to exchange, transfer, lease, or mortgage, led
to an increase in the share of land devoted to long-term crops and labor used in non- agricultural activities (Do & Lyer, 2006). In China, an increase in land tenure security
36
induced more investment in land (Jacoby, Li, & Rozelle, 2002). A land titling program
in Indonesia was found to increase investment incentives and land values (Do & Lyer,
2006; SMERU, 2002). Studies from Latin American countries also show the positive link between land title and various social benefits. Alston, Libecap, and Mueller (1999) found that titles have a positive impact on land investment and land values in some provinces in
Brazil. Galiani and Schargrodsky (2004 and 2010) found that titles have positive impacts on child health, housing investment and house structures in Argentina, whereas Field
(2007) emphasized that formal land titles increased participation in formal labor markets in the urban squats of Peru.
The studies from Africa have drawn researchers’ attention due to their inconclusive results on the effect of land rights. Some studies provide results supportive of models of the farm sector that yield private and social benefits from establishing de jure land rights, whereas others do not. Besley (1995) and Abdulai et al. (2011) found that land rights in Ghana enhance land investment. Another study of Ghana by Goldstein and Udry (2008) found that tenure insecurity decreases investment in land fertility as households who have weak networks of social and political power tend to have their land confiscated when it is left fallow. In Ethiopia, Deininger and Jin (2006) illustrated that improvement in rights to transfer has a positive effect on terrace building and tree planting. In Zambia, Smith (2004) confirmed that land tenure security increases fixed investment and productivity. Hayes, Roth, and Zepeda (1997) found that land tenure security in Gambia enhanced long-term investment which, in turn, improved agricultural outputs. Other studies including Jacoby and Minten (2007), Braselle et al. (2002), Place and Migot-Adholla (1998), Place and Hazell (1993), Migot-Adholla et al. (1991) found
37
weak effects of land rights on agricultural outcomes. Fenske (2011) utilized nine data sets
from West Africa and reassessed some of the previous empirical studies and found that
land tenure is important for fallow and tree planting but less important for manure and
chemical fertilizer. This result implies that land rights are essential for long-term but not short term land investment.
The view that de jure land rights bring about tenure security and agricultural outcomes can be summarized into two factors, which conceptually affect the process of
adoption and implementation of a land titling program and the ability of the program to
provide the expected outcomes. They are (1) the capacity of land governance institutions
and (2) cultural diversity and/or context-specific conditions.
2.2.1. Capacity of land governance institutions
Land governance institutions are complementary factors to de jure land rights that
establish land tenure security and thereby promote investment in land. The capacity of
land governance institutions includes the ability of the state to facilitate land registration
and to enforce the land owner’s rights. Without the above facilitation and protection from
the state, land titles hold little value or utility. Facilitation of land registration requires that land institutions simplify the registration process and that the cost of registering a land transfer with the national land registry does not discourage title holders from registering a land transfer. Facilitation of the initial registry of de facto claims and the
subsequent registry of transfers of de jure land rights means that the land registry will
have up-to-date information on the geographic extent of land claims (often from a survey
of the claim required for initial registration); on restrictions on the land rights, e.g., an
easement for bringing sheep from neighboring farms to market; on previous land owners,
38 the current land owner, and liens on the land. This information allows land markets to function efficiently in placing values on land. Also, for a system of de jure land rights to function properly, there must be a well-functioning legal system capable of adjudicating conflicts over rights and enforcing its decisions (So, 2009). In its absence, households will need to allocate additional resources to guard or fence their properties (Feder &
Feeny, 1991).
The achievements of the land titling program in Thailand highlight the strength of Thailand’s Department of Land and the government’s political will to enhance land tenure security. The role and responsibility of Thailand’s Land Department was clearly defined to ensure that the land titling program was properly implemented
(Rattanabirabongse, Burns, & Nettle, 1998). Institutional capacity has been a major issue for some African countries that have introduced land titling programs (Jacoby & Minten,
2007; Jansen & Roquas, 1998; Platteau, 1996; Attwood, 1990). The major drawback of
Kenya’s land titling program has been the inability of the land governance institution to update records of land transfers (Pinckney & Kimuyu, 1994). Experience from Honduras reveals that its land titling program has had limited effects because the government cannot provide protection for de jure rights holders in a rural community against outside entities. This inability leads to more land conflicts, with de jure rights holders seeking the help of political leaders rather than courts to enforce their rights (Jansen & Roqas, 1998).
Atwood (1990) also highlighted issues related to high transfer costs which discourage households from participating in land titling programs and make them rely on customary or informal land transfers.
39
2.2.2 Cultural diversity and context-specific conditions
Conventionally, land titling programs are enacted to increase land tenure security, enhance agricultural productivity by facilitating transfers of land to more productive managers, increase access of land owners to formal credit institutions, and provide opportunities for land owners to use new technology and to invest more in their farms.
However, land titling programs do not always produce these results. One reason may be that researchers tend to overlook the cultural heritage of each country. As Plateau (1996) suggested, many problems accompanying a developing country’s land titling program stem from a failure to take into account sequences of evolution in land rights. He views the evolution from customary land rights enforced at the village level to de jure individual land rights enforced at the provincial or national level to be due to population pressure, change in technologies, and increases in commercialization of agriculture.
As arable land becomes more scarce in a country with a growing population, people demand rules or institutions compatible with factor endowments and technologies.
A land titling program may be more appropriate in such conditions, as the higher value of the land warrants a more careful specification and more rigorous enforcement of the package of rights and responsibilities adhering to the land (Platteau, 1996). Some studies have shown that introducing a land titling program, especially in African countries where indigenous land rights are still prevalent, may create uncertainty and possibly conflicts rather than increase land tenure (Jacoby & Minten, 2007; Platteau, 1996; Atwood, 1990).
In some African countries, the technology package necessary to make larger investments in land is unavailable, as their agricultural sector is not yet commercialized. Platteau
40
(1996) concluded that land registration in Africa should only be considered when
informal institutions and practices cannot provide tenure security.
2.2.3 Issues in the empirical studies
Inconclusive effects of land rights on agricultural outcomes in empirical studies
may be partly due to flaws in these studies. First, measures of land investment vary
considerably. Land investment in most studies is measured with discrete variables such as
building fences, digging wells, constructing drainage systems, terracing, irrigating,
mulching, applying manure, conserving soil, and planting trees. Some types of
investments may not be initiated until land tenure security is high and the quality of land
is sufficiently high to warrant such investments. Deininger and Jin (2006) warned that the
use of such variables as the “application of manure” to measure land investment does not
provide a clear indication of how the outcomes relating to such small activities translate
into a larger intervention. On the other hand, land investments that are either commonly
or rarely made pose several problems. Fenske (2011) maintained that “excluding them
can, for example, hide insecurity so severe that it has completely discouraged a particular
investment. Lack of identifying variation in the dependent variable raises the standard
errors, pushing results towards insignificance” (p. 142).
Second, de jure land rights are rarely exogenously imposed. Households with more education, wealth, and political connections are more likely to participate in land titling programs. In some countries, land investments facilitated improvements in tenure security for households with fragile de facto land rights (Atwood, 1990). A study of
Burkina Faso land holders confirmed that land investment served to increase land tenure
41
security rather than the other way around (Brasselle et al., 2001). A similar pattern was
found in Uganda where coffee planting was undertaken to enhance tenure security (Place
& Otsuka, 2002). Despite the causal relationships identified between land rights and investment, some empirical studies of this relationship did not control for endogeneity of land rights.1
Third, the issue of soil quality in land investment regression analysis has never
been widely discussed. Conceptually, soil quality is related to the shape and relief of the
plot, the drainage capacity of the soil, and the geographical area of the plot in relation to
irrigation systems or water resources. Even within relatively small geographic areas, there can be large variations in soil quality (Benjamin, 1995). Previous studies reported a small effect of soil quality on land investment (Jacoby & Minten, 2007; Place & Otsuka,
2002; Hayes et al., 1997; Gavian & Fafchamps, 1996; Besley, 1995). A dummy variable
measuring soil quality was used in almost every econometric study but for the Feder et al.
(1988) study of land titling in Thailand. It used a soil quality index incorporating
information on type of soil, slope of the land, upland or lowland location, access to
irrigation, availability of all-weather roads to village and market, and travel time to
nearest market and village. The soil quality index was found to have a positive and
statistically significant impact on land investment (Feder et al., 1988).
Fourth, the finding of a weak relationship between land rights and land
investment in some studies was based on empirical analyses using small samples. A small
sample size may not provide sufficient variation in land rights to allow precise estimates
of this relationship, and there is a tendency for coefficient estimates to be biased toward
1 Brasselle et al. (2001) discussed numerous empirical studies which did not control for endogeneity of land rights. 42 zero (Deininger & Jin, 2006; Deaton, 1997). Besley (1995) also examined this issue in his seminal study, arguing that the statistically insignificant link between land investment and land rights in Anloga (Ghana) could be due to the small sample size utilized in regression estimates.
2.3 Prospects for enhancing land tenure security in Cambodia
2.3.1 Agriculture
Agriculture is the most important economic sector in Cambodia. This sector contributes almost one-third of GDP and employs more than 50 percent of the total labor force in the formal economy. From 2000 to 2008, the GDP of the agricultural sector grew
5.6 percent annually (Yu & Diao, 2011). The steady growth for almost one decade came from agro-industries owned by a small group of wealthy people; small farmers did not gain much from this growth (So, 2009; Kenjiro, 2005). Cambodia’s outputs per hectare of common crops, such as rice, maize, and cassava are the lowest in the region. The average rice yield in 2003 was about 2.15 tons per hectare, whereas in 2003 Vietnam, Thailand, and Lao reached 4.6, 2.45, and 3.31 tons per hectare respectively (World Bank, 2006).
Cambodia’s output per hectare in 2003 just slightly results (about 2 tons) from the 1960s and early 1970s (Sarthi & Chan, 2002).
The low productivity of this sector can be partly attributed to a lack of land tenure security, low investment and limited access to credit institutions (World Bank, 2002).
According to CSES (2003-2004), from 1979 to 2004 approximately 15 percent of agricultural land parcels were invested in (see table 2.2). Loans from formal credit institutions were low (only 3.6 percent of households received them), as most households
43 instead approached money lenders and non-governmental organizations. The average interest rate paid by households was about 4.4 percent per month. Borrowing for agricultural purposes accounted for about 24 percent of household loans, and the rest of the borrowed funds supplemented household consumption, medical spending, or other purposes (see table 2.3). Cultivated land per household did not grow fast enough to feed
Cambodia’s growing population. The population of Cambodia grew from 6 million in the early 1980s, to more than 11 million in 1998, and 13.4 million in 2008 (Chan & Sharti,
2002; Census of Cambodia, 2008). Each household occupied only 1.29 hectares in 2004
(CSES, 2003-2004).
2.3.2 Land investment in Cambodia (1979 to 2004)
This section uses CSES data from 2003-2004 to compare investment in land during the 1979-1988 period when land was under state control, and the 1989-2004 period when land parcels were privatized and de facto and de jure land rights systems coexisted. The separation of our sample into two periods allows observations of investment patterns in land parcels during periods when land parcels were owned by the state and collectively held and when land parcels were privately owned.
Table 2.1 provides the total number of each type of land investment made from
1979 to 2004. There are eight types of land investment: digging a well, digging a ditch, terracing, drainage construction, soil reclamation, growing an orchard, planting a perennial tree, and other investment. The CSES series for land investments began in 1979 when land had not yet been privatized. Column 10 of Table 2.1 provides the total number of investments in all categories. Households made a total of 323 investments in their
44 cultivated lands in 1979, 136 in 1980, 58 in 1981, and 101 in 1982. Land investment in
1979 was higher than in all other years. From 1983 to 1999, the total number of land investments decreased before reversing course beginning in 2000. In both periods (1979-
1988 and 1989-2004), investments were made on about 15 percent of cultivated farms.
Investments in digging ditches, planting trees, and establishing orchards were the most frequent. Investments in terracing were negligible after 1979.2
In the first period (1979 to 1988), cultivated land was collectively held by the solidarity group. Investment was nonetheless quite substantial in 1979 in comparison with other years in this period. Investment after land privatization never reached the 1979 pre-privatization peak of 323 per year. Two arguments for the large investment in land when it has not yet been privatized are considered here. First, households might have invested more because they perceived that they were the legitimate owners of the land and that making investments in the land would solidify their claims. Second, land investment was indispensable due to increases in the demand for food after the Khmer
Rouge Regime (KRR) was driven out of Phnom Penh and the towns and from a certain part of the countryside in 1979. The KR soldiers destroyed whatever food stocks they could not take with them as they retreated, and because farmers had not been allowed to retain their own food under the KRR (So, 2009). Consequently, in 1979 there was a severe shortage of food that threatened rural households’ livelihoods. Thus, after the end of the KRR, households made considerable investments in land regardless of their tenure situation to foster their immediate survival.
In the second period (1989 to 2004), rights to land were individualized formally, based on a series of laws that enacted from 1989. People claimed rights to land either
2 Figure 2.1 shows investment patterns from 1979 to 2004. 45 with some form of documents (e.g., land titles, application receipts, and land investigation documents) or without any documents (a de facto land claim). The increase in the total number of land parcels investment in this period, from 54 in 1989 to 137 in
2000, to 118 in 2001, 119 in 2002, and 137 in 2003 could have been due to an increase in land tenure security or the belief that investment would increase land tenure security. As discussed in Chapter One, prior to the adoption of the 2001 land law, households were eligible to claim rights on the plots they had held peacefully for five consecutive years.
Thus it remains unclear whether the increase in investment during the second period was because of an increase in tenure security or because of the belief that investment in a land parcel strengthens the land tenure of occupants with de facto rights.
2.4 A model of tenure security and investment
Feder et al. (1988), Besley (1995), Deininger and Jin (2006) developed theoretical models to assess the impact of tenure security on choices of farmers. The Feder et al.
(1988) model derives relationships between land rights and the price of land, land investment, and access to credit markets. Besley (1995) develops three theoretical models to explain the link between land property rights and investment decisions. He posits that having secured land rights would increase investment in land in three scenarios: land rights reduce the risk of expropriation; land rights allow land to be used as collateral in investment markets; land rights facilitate sale or rent to households with a higher value of the land. Deininger and Jin (2006) used a two-period household model to explain the effect of tenure security on investment. They suggested that tenure security will have a positive impact on investment by inducing a productivity increase if land tenure is
46
exogenously given. This result does not hold, however, when farmers invest to enhance
land tenure security.
The three models are similar in positing that land tenure security encourages land investment. However, the model developed by Deininger and Jin (2006) fits the
Cambodia de jure property rights’ context, as discussed in Chapter One. The model provided below is taken from Deininger and Jin (2006). A Cobb-Douglas production
function is used to simplify discussion. Households are assumed to be endowed with the
same amount of land and to maximize their utility over two periods:
( 1, 2) = 1 + 2
𝑀𝑀𝑀𝑀𝑀𝑀s.t. 𝑈𝑈 𝐶𝐶 𝐶𝐶 𝑙𝑙𝑙𝑙𝐶𝐶 𝛽𝛽𝛽𝛽𝛽𝛽𝐶𝐶
1 (1 + )[( 1) ( 1) + 1 1 1] 𝛼𝛼 𝑃𝑃 −𝛼𝛼 𝑂𝑂 1 + [ (𝑟𝑟2( 𝐾𝐾1, 1))𝐿𝐿( 1 + (𝐿𝐿1)𝑤𝑤) (− 2𝐶𝐶) + 2 2 2] = 0 𝐼𝐼 𝐼𝐼 𝛼𝛼 𝑃𝑃 −𝛼𝛼 𝑂𝑂 1 +𝜏𝜏 𝑆𝑆1 +𝑆𝑆 𝐿𝐿1 = 𝐾𝐾1 and𝑔𝑔 2𝐿𝐿+ 2 =𝐿𝐿 2 𝐿𝐿 𝑤𝑤 − 𝐶𝐶 𝐼𝐼 𝑂𝑂 𝑃𝑃 𝑂𝑂 𝑃𝑃 where: 𝐿𝐿 𝐿𝐿 𝐿𝐿 𝐿𝐿 𝐿𝐿 𝐿𝐿 𝐿𝐿
1and 2= Consumption in periods 1 and 2
𝐶𝐶1and 𝐶𝐶2= Labor in periods 1 and 2
𝐿𝐿1, 1,𝐿𝐿 and 1 = Labor used for agricultural production, land related investment, 𝑃𝑃 𝐼𝐼 𝑂𝑂 and𝐿𝐿 𝐿𝐿off-farm𝐿𝐿 work with wage w.
1= Capital in period 1
1 𝐾𝐾1 = Output in period 1 which is generated according to ( 1) ( 1) 𝛼𝛼 𝑃𝑃 −𝛼𝛼 𝑌𝑌2= 1 + ( 1), ( 1) is an increasing function of 1 𝐾𝐾 𝐿𝐿 𝐼𝐼 𝐼𝐼 𝐼𝐼 1 𝐾𝐾2= Output𝐾𝐾 𝑔𝑔 generated𝐿𝐿 𝑔𝑔 𝐿𝐿 in period 2 according to ( 1 +𝐿𝐿 ( 1)) ( 2) 𝐼𝐼 𝛼𝛼 𝑃𝑃 −𝛼𝛼 𝑌𝑌 [0,1]= Discount rate of Cambodian farmers𝐾𝐾 𝑔𝑔 𝐿𝐿 𝐿𝐿
𝛽𝛽 ∈ 47
[0,1]= Probability that land can be seized in period 2
𝜏𝜏1∈ and 2= Land tenure security in periods 1 and 2.
Let𝑆𝑆 [ 𝑆𝑆2( 1, 1)] “be a function of land tenure security in the second period”, 𝐼𝐼 which is a 𝜏𝜏function𝑆𝑆 𝑆𝑆 𝐿𝐿 of tenure security and land improvement in the first period and
2 2 2 2 assume that 2 > 0, 2 < 0, 2 1 > 0, 2 1 < 0, 2 1 > 𝐼𝐼 2 2 2 0, 2 𝜕𝜕1𝜕𝜕<⁄𝜕𝜕0𝑆𝑆, and 𝜕𝜕2 𝜏𝜏⁄𝜕𝜕1 𝑆𝑆 1 < 0 𝜕𝜕(Deininger𝑆𝑆 ⁄𝜕𝜕 𝑆𝑆 & Jin,𝜕𝜕 𝑆𝑆 2006,⁄𝜕𝜕 𝑆𝑆p. 1270).𝜕𝜕 𝑆𝑆Assuming⁄𝜕𝜕 𝐿𝐿 𝐼𝐼 𝐼𝐼 separability𝜕𝜕 𝑆𝑆 ⁄𝜕𝜕 𝐿𝐿between household𝜕𝜕 𝑆𝑆 ⁄𝜕𝜕𝑆𝑆 consumption𝜕𝜕 𝐿𝐿 and production, the household utility
maximization problem above can be written as:
1 (1 + )[( 1) ( 1) + 1 1] , , , , 1 1 1 2 2 𝛼𝛼 𝑃𝑃 −𝛼𝛼 𝑂𝑂 𝐼𝐼 𝑃𝑃 𝑂𝑂 𝑃𝑃 𝑂𝑂 𝐿𝐿 𝐿𝐿 𝑀𝑀𝑀𝑀𝑀𝑀𝐿𝐿 𝐿𝐿 𝐿𝐿 𝑟𝑟 𝐾𝐾 𝐿𝐿 𝐿𝐿 𝑤𝑤 1 + [ ( 2( 1, 1))( 1 + ( 1)) ( 2) + 2 2] 𝐼𝐼 𝐼𝐼 𝛼𝛼 𝑃𝑃 −𝛼𝛼 𝑂𝑂 s.t𝛽𝛽. 𝜏𝜏 𝑆𝑆 𝑆𝑆 𝐿𝐿 𝐾𝐾 𝑔𝑔 𝐿𝐿 𝐿𝐿 𝐿𝐿 𝑤𝑤
1 + 1 + 1 = 1 and 2 + 2 = 2 𝐼𝐼 𝑂𝑂 𝑃𝑃 𝑂𝑂 𝑃𝑃 The first𝐿𝐿-order𝐿𝐿 conditions𝐿𝐿 𝐿𝐿 are: 𝐿𝐿 𝐿𝐿 𝐿𝐿
= (1 + )(1 )( 1) ( 1 1 1 ) 1 𝜕𝜕𝜕𝜕 𝑂𝑂 𝐼𝐼 𝛼𝛼 𝐼𝐼 −𝛼𝛼 𝜕𝜕𝐿𝐿 − 𝑟𝑟 − 𝛼𝛼 𝐾𝐾 𝐿𝐿 − 𝐿𝐿 − 𝐿𝐿 1 + ′ 2( 1, 1) 2′( 1, 1)( 1 + ( 1)) ( 2) 𝐼𝐼 𝐼𝐼 𝐼𝐼 𝛼𝛼 𝑃𝑃 −𝛼𝛼 𝛽𝛽�𝜏𝜏 �𝑆𝑆 𝑆𝑆 𝐿𝐿 �𝑆𝑆 𝑆𝑆 𝐿𝐿 𝐾𝐾 1𝑔𝑔 𝐿𝐿 1 𝐿𝐿′ � + [ ( 2( 1, 1))( 1 + ( 1)) ( 2) ( 1) = 0 (1) 𝐼𝐼 𝐼𝐼 𝛼𝛼− 𝑃𝑃 −𝛼𝛼 𝐼𝐼 𝛽𝛽 𝛼𝛼𝛼𝛼 𝑆𝑆 𝑆𝑆 𝐿𝐿 𝐾𝐾 𝑔𝑔 𝐿𝐿 𝐿𝐿 𝑔𝑔 𝐿𝐿 = (1 )( 1) ( 1 1 1 ) = 1 (2) 𝜕𝜕𝜕𝜕1 𝑂𝑂 𝛼𝛼 𝐼𝐼 𝑂𝑂 −𝛼𝛼 𝜕𝜕𝐿𝐿 − 𝛼𝛼 𝐾𝐾 𝐿𝐿 − 𝐿𝐿 − 𝐿𝐿 𝑤𝑤 = (1 )( 1 + ( 1)) ( 2) = 2 (3) 𝜕𝜕𝜕𝜕2 𝑂𝑂 𝐼𝐼 𝛼𝛼 𝑃𝑃 −𝛼𝛼 𝜕𝜕𝐿𝐿 − 𝛼𝛼 𝐾𝐾 𝑔𝑔 𝐿𝐿 𝐿𝐿 𝑤𝑤 Equations (2) and (3) imply that the marginal product of labor is equal to the off- farm wage, whereas equation (1) implies that the marginal cost of investment effort in period one is equal to the discounted value of the marginal benefit of the increase in
48 tenure security in the second period and the marginal benefit of the additional investment chosen in period two.
2.4.1 Case 1: Land tenure security is exogenous
If land tenure security is exogenous, the second term in equation (1) drops out.
Equation (1) then can be written as:
= (1 + )(1 )( 1) ( 1 1 1 ) 1 𝜕𝜕𝜕𝜕 𝑂𝑂 𝐼𝐼 𝛼𝛼 𝐼𝐼 −𝛼𝛼 𝜕𝜕𝐿𝐿 − 𝑟𝑟 − 𝛼𝛼 𝐾𝐾 𝐿𝐿 − 𝐿𝐿 − 1𝐿𝐿 1 ′ + [ ( 2( 1, 1))( 1 + ( 1)) ( 2) ( 1) = 0 (4) 𝐼𝐼 𝐼𝐼 𝛼𝛼− 𝑃𝑃 −𝛼𝛼 𝐼𝐼
If tenure security𝛽𝛽 is𝛼𝛼 𝛼𝛼exogenous,𝑆𝑆 𝑆𝑆 𝐿𝐿 2𝐾𝐾 1, 1𝑔𝑔 can𝐿𝐿 be replaced𝐿𝐿 with𝑔𝑔 𝐿𝐿 1. Equation (1) then 𝐼𝐼 becomes: 𝑆𝑆 �𝑆𝑆 𝐿𝐿 � 𝑆𝑆
= (1 + )(1 )( 1) ( 1 1 1 ) 1 𝜕𝜕𝜕𝜕 𝑂𝑂 𝐼𝐼 𝛼𝛼 𝐼𝐼 −𝛼𝛼 𝜕𝜕𝐿𝐿 − 𝑟𝑟 − 𝛼𝛼 𝐾𝐾 𝐿𝐿 1− 𝐿𝐿 −1 𝐿𝐿 ′ + [ ( 1)( 1 + ( 1)) ( 2) ( 1) = 0 (5) 𝐼𝐼 𝛼𝛼− 𝑃𝑃 −𝛼𝛼 𝐼𝐼 Substituting equations𝛽𝛽 𝛼𝛼𝛼𝛼 𝑆𝑆 (2) 𝐾𝐾and (3)𝑔𝑔 𝐿𝐿into equation𝐿𝐿 (5):𝑔𝑔 𝐿𝐿
′ = (1 + )(1 ) 1 + [ ( 1) 2 ( 1)] = 0 (6) 1 𝜕𝜕𝜕𝜕 𝐼𝐼 𝐼𝐼 𝜕𝜕𝐿𝐿 − 𝑟𝑟 − 𝛼𝛼 𝑤𝑤 𝛽𝛽 𝛼𝛼𝛼𝛼 𝑆𝑆 𝑤𝑤 𝑔𝑔 𝐿𝐿 Totally differentiating (6) with respect to 1 and 1: 𝐼𝐼 ′ ′ [ ( 1) 2 ′( 1)] 1 + [ ′(𝐿𝐿1) 2 𝑆𝑆( 1)] 1 = 0 (7) 𝐼𝐼 𝐼𝐼 𝐼𝐼 𝛽𝛽 𝛼𝛼𝛼𝛼 𝑆𝑆 𝑤𝑤′( 𝑔𝑔) ′ 𝐿𝐿 𝜕𝜕𝐿𝐿 𝛽𝛽 𝛼𝛼𝛼𝛼 𝑆𝑆 𝑤𝑤 𝑔𝑔 𝐿𝐿 𝜕𝜕𝑆𝑆 1 = 1 1 > 0 (8) 𝐼𝐼 ( ) ′′ 𝐼𝐼 𝜕𝜕𝐿𝐿1 𝜏𝜏 𝑆𝑆1 𝑔𝑔 �𝐿𝐿1� 𝐼𝐼 𝜕𝜕𝑆𝑆 − 𝜏𝜏 𝑆𝑆 𝑔𝑔 �𝐿𝐿 � The last result indicates that land tenure security has a positive impact on land investment. This is the standard case in the literature: Tenure security enhances land investment which increases land productivity.
49
2.4.2 Case 2: Land tenure is endogenous
If a household invests to enhance land tenure security, then the third term of
equation (1) drops out:
= (1 + )(1 )( 1) ( 1 1 1 ) 1 𝜕𝜕𝜕𝜕 𝑂𝑂 𝐼𝐼 𝛼𝛼 𝐼𝐼 −𝛼𝛼 𝜕𝜕𝐿𝐿 − 𝑟𝑟 − 𝛼𝛼 𝐾𝐾 𝐿𝐿 − 𝐿𝐿 − 𝐿𝐿 1 + ′ 2( 1, 1) 2′( 1, 1)( 1 + ( 1)) ( 2) = 0 𝐼𝐼 𝐼𝐼 𝐼𝐼 𝛼𝛼 𝑃𝑃 −𝛼𝛼 = (𝛽𝛽1�𝜏𝜏+�𝑆𝑆) 𝑆𝑆 +𝐿𝐿 �𝑆𝑆 𝑆𝑆( 𝐿𝐿, 𝐾𝐾) (𝑔𝑔 𝐿𝐿, ) [ 𝐿𝐿 2 �] = 0 1 ′ 2 1 1 2′ 1 1 (1 ) (9) 𝜕𝜕𝜕𝜕1 𝑤𝑤 2 𝐼𝐼 𝐼𝐼 𝐼𝐼 𝑃𝑃 𝜕𝜕𝐿𝐿 − 𝑟𝑟 𝑤𝑤 𝛽𝛽�𝜏𝜏 �𝑆𝑆 𝑆𝑆 𝐿𝐿 �𝑆𝑆 𝑆𝑆 𝐿𝐿 � −𝛼𝛼 �𝐿𝐿 � Totally differentiating (9) with respect to 1 and 1: 𝐼𝐼 = {[ "( ) 2 + 2 ] 𝐿𝐿 ( , 𝑆𝑆) [ 2 ]} 2 1 1 2′ 1 1 (1 ) 𝜕𝜕𝑆𝑆1 𝜕𝜕𝑆𝑆1 𝑤𝑤 2 𝐼𝐼 𝐼𝐼 𝐼𝐼 𝑃𝑃 𝜕𝜕𝑆𝑆 𝜕𝜕𝐿𝐿 −𝛼𝛼 �𝐿𝐿 � 𝜏𝜏 𝑆𝑆 � 𝜕𝜕𝑆𝑆 𝜕𝜕𝐿𝐿 � 𝑆𝑆2 𝑆𝑆 𝐿𝐿 2 2 2 2 +{ ′ 2( 1, 1) 1 + 2 1 } = 0 (1 ) 2 1 1 1 𝐼𝐼 𝑤𝑤 𝜕𝜕 𝑆𝑆 𝜕𝜕 𝑆𝑆 𝐼𝐼 𝑃𝑃 𝐼𝐼 𝐼𝐼 𝜏𝜏 �𝑆𝑆 𝑆𝑆 𝐿𝐿 � � −𝛼𝛼 �𝐿𝐿 �� �𝜕𝜕𝐿𝐿 𝜕𝜕𝑆𝑆 2 𝜕𝜕𝑆𝑆 𝜕𝜕 𝐿𝐿 𝜕𝜕𝐿𝐿 � 2 2 [ "( 2) 2′ 1, ]+[ ′ 2 1, ] 1 1 1 𝜕𝜕𝑆𝑆 𝜕𝜕1𝑆𝑆 1 = 𝜏𝜏 𝑆𝑆 𝑆𝑆 �𝑆𝑆 𝐿𝐿𝐼𝐼 � 𝜏𝜏 �𝑆𝑆 �𝑆𝑆 𝐿𝐿𝐼𝐼 �� 2 < 0 (10) 1 "( 𝜕𝜕)𝑆𝑆 2 ′ , + ′ , 𝐼𝐼 2 𝜕𝜕𝜕𝜕 2 2 1 1 2 1 1 𝜕𝜕2𝐿𝐿 𝜕𝜕𝑆𝑆 𝐼𝐼 𝜕𝜕𝑆𝑆1 𝜕𝜕 𝑆𝑆1 𝜕𝜕𝐿𝐿 𝜏𝜏 𝑆𝑆 𝑆𝑆 �𝑆𝑆 𝐿𝐿𝐼𝐼 � 𝜏𝜏 �𝑆𝑆 �𝑆𝑆 𝐿𝐿𝐼𝐼 �� − 𝜕𝜕𝐿𝐿𝐼𝐼 𝜕𝜕 𝐿𝐿𝐼𝐼 In this case, land rights are endogenous, as households invest to enhance land
tenure security. Investments to enhance security are mostly short-term investments and
some of these investments, such as posting or fencing, are not made to increase
productivity but rather to secure land rights.
2.5 Empirical analysis
2.5.1 Data
The empirical analysis is based on a Cambodia Socio-Economic Survey (CSES) conducted in 2003-2004 by the National Institute of Statistics (NIS) of the Ministry of
Planning (MOP). The survey was funded by the United Nations Development Program,
Swedish International Development Agency, and the World Bank. NIS is an experienced
50
institution that administers most of the national surveys in Cambodia. The first survey
was carried out in 1994 and CSES 2003-2004 is the fifth survey. It covers 100 percent of
the sample population, which compares well to the other four surveys which left some
geographic areas uncovered due to poor security in some areas. The goal of CSES is to
collect information on various socio-economic aspects of households to track down the extent of household poverty.
It is a nationally representative survey of 15,000 households in 24 provinces and
900 villages conducted from November 2003 to January 2005. The sampling frame of the survey was based on Cambodia’s general population census. Selected households were contacted to fill out the first part of the questionnaire covering household composition. Households were then obligated to fill out a diary of expenditure and income. Visits to and interviews of the households continued until all sections of the questionnaire were completed. The survey is available at individual, household, and village levels. The individual and household datasets contain information on employment, education, consumption and income, child health, housing construction, and farmland. The village dataset contains general information on population, agricultural land, development projects, schooling, presence of a hospital, common property resources, and occurrence of natural disasters.
2.5.2 Farm characteristics
The survey asked households to report the number of farmlands and type of documents connected with each farmland. The types of documents are land title, land investigation document, application receipt, rental contract, and other specific documents.
51
Generally, households with land titles are protected under the law and are entitled to transfer rights. Holders with other documents such as an application receipt or a land investigation document have to go through several steps (reviewed in Chapter One) to obtain a title. To some extent, people face a higher risk of being expropriated if their lands are registered by others; those who hold no documents are in even greater danger.
Table 2.2 provides the mean and standard deviation for plots by type of document, mode of land acquisition, land investment, and land type. General information regarding plot characteristics shows that the majority of plots are held without formal documents, are non-irrigated plots with only a few land investments, and are a type of wetland. Table 2.2 indicates that among plots reported held with some type of document,
21 percent are held with land titles, 24 percent with application receipts, 3 percent with various other documents, and 52 percent without any documents.3 Most households either obtained land from the government (45 percent) or inherited land from their families (32 percent). Only 12 percent of land was purchased. Approximately 2 percent of plots were involved in litigation. The length of occupation of each plot was about 18 years on average. 77 percent of plots are wetlands and 75 percent are non-irrigated plots.
Together this implies that cultivation in Cambodia remains rainfed, and outputs depend heavily on weather conditions. Information on type of plot also tells the story about soil quality. Land investment such as digging a well or a ditch, construction of drainage systems, and planting trees has been conducted only at minimal levels. Only 15 percent of the plots have been invested in, with 5 percent invested in digging a well or a ditch, 5
3 As mentioned in Chapter One, the percentage of titled plots is much lower in the CSES 2009. The higher percentage of titled plots in CSES 2004 may be due to confusion between land title and application receipt, as some households have treated their application receipt as a land title certificate. 52
percent in drainage construction, 3 percent in growing orchards, and 2 percent in planting
trees. Other types of investments have been neglected.
This Chapter’s analysis uses three different measures of land rights: land titles, land titles or application receipts, and application receipts. The land rights variables used in this study are binary. As discussed in Chapter One, some households treat application receipts as land titles. Although legally they are not, some households believe they have at least some security of tenure with an application receipt. So (2009) illustrated that some villagers who have not yet completed their registrations indicate that they are not worried about expropriation as long as they stay on the land and keep it productive. For this reason, households who hold farms with land titles or application receipts could potentially have the same or similar investment behaviors. Therefore, it is important to look at the effects separately and to estimate the magnitude of the impact with different measures of land rights.
Land investment is measured by the type of long-term investment conducted by households after the 1989 reform. The survey includes three different types of land investment. Because the second and third types are relatively unimportant, I do not consider them in this study. The first type measures long-term investment in land and uses binary variables to measure household investment in more narrow categories. The variables consist of planting trees, preparing and planting orchards, digging ditches and wells, constructing drainages and terraces, and taking measures to conserve soil. For this study, the different categories of type one investments are combined due to a small number of investments in each category.
53
2.5.3 Household characteristics
Table 2.4 shows that on average household heads have only 4.37 years of
education and do not complete primary school. The average household has 4.94 persons,
more female than male members, and is headed by a male (female heads only account for
22 percent of household heads) who is 45.14 years old. The average land holding of each
household is 1.29 hectares held over 1.31 plots. The average distance from a household in
each village to an all-weather road is approximately 3 kilometers. Information on
household income is not included in Table 2.4, as household income in developing
countries has large annual variations and is very imprecisely measured because
households often hide information on their real income (Deaton, 1997). Instead, table 2.4 includes information on the value of durable goods, livestock, and number of rooms in the farmer’s house as measures of household wealth. These measures are also used in
Besley (1995).
2.5.4 Empirical models and econometric issues
The theoretical model discussed above suggests a regression of land investment for plot j, operated by household i, on land rights, and household and plot characteristics:
= + + + (11)
𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑗𝑗 𝑖𝑖 𝑖𝑖𝑖𝑖 where 𝐼𝐼 represents𝛽𝛽𝑆𝑆 𝛼𝛼household𝑃𝑃 𝜑𝜑𝑍𝑍 i 𝜀𝜀investment on plot j, is a binary variable denoting
𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 whether𝐼𝐼 household i holds land with title or application𝑆𝑆 receipt, indicates plot j
𝑗𝑗 characteristics, and denotes household i characteristics. The specification𝑃𝑃 of the
𝑖𝑖 investment equation 𝑍𝑍will be more concrete with the inclusion of available measures of
household and plot characteristics. Heterogeneity across households is important in
54 explaining household investment decisions. Furthermore, the variations in quality and location across plots also have an impact on land investment as well as households’ efforts to enhance their rights to plots (Besley, 1995). Thus, the empirical analysis of land investment in this study includes all available variables measuring plot characteristics from Table 2.2 and household characteristics from Table 2.4. Soil quality is proxied by type of land, and access to an all-weather road is used to measure location and distance of plots. Normally, the plots are not far from household residences.
Land rights could, of course, be endogenous, such that land tenure security in period two is related to tenure security and land related-investment in period one. Failure to take either measurement error in land rights into account or to model the relationship correctly tends to bias estimates of the impact of land rights on land investment (Besley,
1995). Most previous econometric studies of this relationship used instrumental variable techniques to account for endogeneity of land rights. A few recent studies used matching regression techniques to account for this selection bias (Galiani & Schargrodsky, 2010;
Field & Torero, 2006; Galiani & Schargrodsky, 2004).
There are several sensible instrumental variables for land rights that are well documented by the previous microeconomic studies. These are rights to transfer the plot, litigation over the plot, length of plot occupation, and mode of land acquisition
(Markussen, 2008; Besley, 1995). While these instrumental variables are available in the
CSES surveys, the analysis of land investment in this study uses the PSM method instead.
The data support the use of this method, and below I discuss several advantages to using the PSM technique.
55
2.5.5 Propensity score matching (PSM)
The PSM method, initially proposed by Rosenbaum and Rubin (1983), has gained popularity in evaluating employment, education, and health programs (Caliendo &
Kopeini, 2008). The method has been viewed as an alternative to experimental methods
in the field of employment program evaluation if certain conditions are fulfilled (Peikes,
Moreno, & Orzol, 2008; Heckman, Ichimura, Smith, & Todd, 1998). PSM is different
from other matching methods because it produces a single score that summaries multiple
characteristics. Its main advantage is the reduction of matching dimensionality (Peikes et
al., 2008).
Despite its popularity, some researchers have warned of limitations. First, using
this method requires “knowledge of the experimental impact estimates” to assess whether
it produces unbiased estimates. Second, it requires “data on the right variables that would
accurately predict program participation” (Peikes et al., 2008, p. 222; Smith & Todd,
2005). This method is data demanding and interpretation of results based on small
samples of non-participants should be cautiously made because the process of matching
requires the treatment and control group members with the same propensity scores to
balance with respect to distributions of covariates covering the treatment and control
samples. For this property to hold, a decent sample size is needed at each value of the
propensity score. Thus, empirical studies, which rely on small samples, potentially bias
the results or increase sampling errors (Bryson, Dorsett, & Purdon, 2002; Zhoa, 2000).
Third, PSM estimation relies on the Conditional Independence Assumption (CIA) which
cannot be tested. Thus, it is hard for researchers to justify whether they include all of the
variables influencing participation and outcome (Bryson et al., 2002).
56
This study uses PSM to account for selection effects in household participation in
Cambodia’s program to convert de facto land claims into formal de jure land rights supported by a national land registry. The technique has not been widely used in land title program evaluation despite its popularity in program evaluations. The CSES survey utilized in this study allows the use of this method. The CSES contains information on most of the variables required to estimate the impact except soil quality, which is not included in the survey. However, CSES includes information on types of land, and this indicator can be used as a proxy for soil quality. The use of a soil quality proxy variable is common in other studies of property rights and land investment. Further, the data also contain sufficiently large samples of treatment and control households to perform matching.
2.5.6 Estimation of propensity scores
Estimation of propensity scores is a crucial task. Choice of model and variable to include in the logit selection model has to be carefully made (Caliendo & Kopeini,
2008).4 For the choice of variables to include in the model, researchers should pay attention to relevant economic theory, previous empirical findings, and information about the institutional setting. Econometricians recommend that only variables that are unaffected by treatment assignment should be included in the model as matching strategies based on CIA. However, more variables should be included in the model when there is uncertainty about model specification (Caliendo & Kopeini, 2008; Smith & Todd,
2005; Sianeis 2004). Bryson et al. (2002) pointed out that inclusion of too many variables
4 Here the choice of model to estimate the probability of participation is not an issue, as treatment assignment is a dichotomous variable, and using either logit or probit would yield the same results. 57 could aggravate the common support problem as well as increase estimated standard errors.
The variables included in the propensity score estimates in this study are selected in the context of the empirical literature discussed in section 2.2. Since this analysis uses three different measures of land rights, the discussion of the results is divided into three panels. Panel A consists of estimated outcomes when a land title is used as a measure of secure land rights. Panel B provides the results when a combined measure of land title and application receipt is used. Panel C gives the outcomes when the application receipt is used. Table 2.5a, 2.6a, and 2.7a present estimated propensity scores derived from logit regressions of titled farms, a combined category, and application receipt farms.
The following analysis focuses on results from an estimation using land title as a measure of land rights. The estimated coefficient on human capital assets is positively associated with the propensity to obtain land title. One more year of education for the household head increases the chance of obtaining land title by 8.8 percent and is statistically significant at the 1 percent level. The ability to secure land title increases with age of the household head, with each additional year enhancing the propensity to obtain title by 2.74 percent (statistically significant at the 10 percent level). The results related to gender of household reflect the role of women in Cambodia as well as their contribution in securing land title. The estimated coefficient on number of female members in the household is positive. One more female member increases the propensity to obtain title by 5.13 percent and is statistically significant at the 5 percent level.
Currently, in Cambodia, many young females from rural areas have migrated to the city or abroad to help relieve their rural household’s burdens. Remittances from these women
58
might increase household consumption and upgrade living standards. This also has a
positive effect on the household’s propensity to obtain land title.
The estimated coefficients on the variables used to measure household wealth
(value of durable goods and number of rooms in each household) are both positive. The
estimated coefficient on the value of durable goods is statistically significant at the 1
percent level, but is very small. One more room in the farmer’s house increases the
probability of obtaining land title by 6.34 percent and is statistically significant at the 5
percent level. The estimated coefficient on the value of livestock is negative and
statistically insignificant. The estimated coefficients on soil quality and access to an all-
weather road are positive. Access to an all-weather road increases by 6.66 percent the
probability of obtaining title and is statistically significant at the 1 percent level. The
estimated coefficient on the size of land operated is not statistically significant.
The outcomes from regressions using application receipt and a combined land
rights variable are mostly the same as those from the regression using land title as the measure of land rights. Two exceptions stand out. First, the estimated coefficient on the
value of livestock becomes statistically significant at the 1 percent level when a
combined land rights measure is used. In the study by Besley (1995), the estimated
coefficient on value of livestock is also statistically significant but the impact is very
small. Second, the estimated coefficient on size of land operated is negative and
statistically significant at the 1 percent level when the application receipt is used as a
measure of land rights. Outcomes vary across regressions for the estimated coefficients
on various measures of soil quality.
59
Figures 2.2, 2.3, and 2.4 provide a histogram summarizing the estimated
propensity scores of treatment and control groups. Visual comparison of the overlap
between treatment and control groups in each bin provides one indication that the overlap
is high enough to perform matching. In the next section, I discuss additional measures to
tackle problems encountered with non-overlapping observations.
2.5.7 Estimation of the effect of land rights through PSM
Matching can be performed through nearest neighbor, radius, kernel, and
stratification techniques to estimate the effect of de jure land rights on investment. Each
method should provide the same outcome as the sample size becomes increasingly large
(Bryson et al., 2002). The choice of matching techniques is crucial, however, when the
sample size is small, as there is a tradeoff between bias and efficiency (Heckman,
Ichimura, & Todd, 1997). Each method has its pros and cons and selection depends
heavily on the nature of the data.
Nearest neighbor matching method involves matching treatment and control
individuals who are closest in terms of their propensity scores. It can be performed with
replacement and without replacement. Matching with replacement refers to multiple uses
of untreated individuals, whereas in the latter case untreated individuals can only be
matched with one treated individual. Allowing for matching with replacement increases
the average quality of match and decreases bias. However, the efficiency of this method
deteriorates as the distance (in estimated propensity scores) to the closest neighbor
increases (Caliendo & Kopeinig, 2008).
60
The radius matching method allows for matches with the nearest neighbor within each caliper as well as all control individuals within the calipers. Individuals outside of the radius are dropped (Bryson et al., 2002). This method presents an opportunity to use as many control individuals within the calipers as possible and therefore allows additional information on good matches to be incorporated in the propensity score estimates. The pitfall of the radius method is that results are very sensitive to the width of the radius selected to perform the matching. The smaller the size of the radius, the more difficult it is to find a match within the range. Consequently, a large number of individuals may not pass the common support requirement (Caliendo & Kopeinig, 2008).
The stratification matching method divides the common support of propensity scores into a set of intervals. The effect is calculated by taking the mean difference in the outcomes between treatment and control individuals within each interval (Caliendo &
Kopeinig, 2008; Smith & Todd, 2005). Kernel matching is a non-parametric matching method. It utilizes the weighted average of all individuals in the control group to create a counterfactual outcome. The advantage of this method is the lower variance which is achieved by using more information. One pitfall is that this method is prone to using observations that are bad matches (Caliendo & Kopeinig, 2008).
The analysis here uses more than one method to test whether the outcome is sensitive to the different methods applied. It discards all investment conducted before
1989 since its purpose is to look at the impact after the reform in 1989. Further, common support is enforced to improve the quality of matching as advised by Becker and Ichino
(2002). A trimming method, used by Dehejia and Wahba (2002 and 1999), is also applied in this study. Comparison group propensity scores which lie below the minimum and
61
above the maximum of estimated scores in the treatment group are discarded. Trimming
leads to 30 observations in the control group being excluded from the sample. The
matching process in this study closely follows the guidelines set forth in Becker and
Ichino (2002). When the balancing property is satisfied, then the matching process can be
performed.
Checking whether the matching procedure is able to balance characteristics
across treatment and comparison groups is suggested as the estimation is not conditioned
on all covariates, rather on the propensity scores. The distribution of propensity scores in
both groups should be similar. I focus on two measures of balance: t-test and bias. The t-
test is used to check if there are significant differences in the covariate means for
treatment and control groups. Table 2.5c, 2.6c, and 2.7c provide comparisons of the
means of treated and control samples before and after matching. Before matching, there are differences in some covariate means but after matching the covariates are balanced in both groups.
Imbens and Wooldridge (2009) indicate that “the difference in averages by treatment status, scaled by the square root of the sum of the variances” is a better method to check for covariate balancing (Imbens & Wooldridge, 2009, p. 24). As suggested in
Imbens and Rubin (2010), a rule of thumb is that the scaled mean difference should be less than 0.25. A larger number indicates that linear regression is sensitive to changes in specification. For all three measures of land rights, the normalized difference in averages
for the matched and control covariates is less than 0.25. In general, the two measures of
balance suggest the covariates for treated and control matched samples are balanced after
matching.
62
Results are reported in Tables 2.5b, 2.6b, and 2.7b. First, the effect of land title on
land investment is discussed. The estimated treatment effect of de jure land rights on land
investment from propensity score matching suggests that, in line with modern theories of
property rights, rights matter for land investment. Investment is about 1.5-1.7 percentage points higher for titled farms and the estimated coefficient on land rights is significant at the 5 percent level. The magnitudes of the effect vary when different matching methods are applied. However, in each method applied, the effect of land rights is positive and remains statistically significant.
When a combined measure is utilized, the results remain positive and statistically significant. Investment is about 1-1.1 percentage points higher for titled farms. The different magnitudes could be due to the number of controls used in performing matching, as some methods discard more observations in the group than other. As articulated earlier, each method has its pros and cons and depends on the nature of the data set. The outcomes from using application receipt as a measure for land rights remain positive in all matching methods applied but the estimated coefficient on land rights is not statistically significant at the 10 percent level.
The results above indicate that land tenure security is important for land investment. When land title or a combined measure of land title and application receipts is utilized, the estimated coefficient on the land rights variable is positive and statistically significant. When application receipts are used to measure property rights in land, the estimated coefficient is not statistically significant at the 10 percent level.
I check the sensitivity of the estimated results as estimation was made under the standard CIA or selection on observables assumption. It is a strong indentifying
63
assumption. A hidden bias may occur if individuals with the same observed covariates
have different chances of receiving treatment. With this hidden bias, results are not robust
(Caliendo & Kopeinig, 2008; Rosenbaum, 2002). Rosenbaum (2002) suggested using a bounding approach to address this problem since the magnitude of selection bias cannot be estimated with non-experimental data. This approach does not test the
unconfoundedness assumption but provides the magnitude of the effect to which any
results depend on this assumption (Becker & Caliendo, 2007, p. 72). Aakvik (2001) and
Becker and Caliendo (2007) recommended using the Mantal and Haenszel test statistic if
the outcome is a binary variable. According to Becker & Caliendo (2007), under the
assumption of no hidden or unobserved selection bias, the odds radio is one (Γ=1). It
implies that the unobserved variables have no impact on the probability of participation.
If the odds radio equals two, it implies that individuals with the same observed covariates
“could differ in their odds of receiving treatment by as much as a factor of two” (Becker
& Caliendo, 2007, p. 73; Rosenbaum, 2002). There are two bounds in the output table:
the Qmh+ adjusts the MH test statistic downward for the case of positive unobserved
selection and Qmh- for the case of negative unobserved selection. Under the assumption
of no hidden bias (Γ=1), the Qmh test statistic is not significant at the 10 percent level for
all measures of land rights. The results from this sensitivity analysis do not imply that de
jure land rights have no effect on land investment. They only indicate that the findings
are “sensitive to possible deviations from identifying unconfoundedness assumption” and
it does not imply that CIA does not hold (Becker & Caliendo, 2007, p. 81).
The results reported in this chapter are smaller than those reported by other
studies that found a positive link between land rights and investment. Besley (1995)
64 found that land rights in Ghana enhanced investment by 2.5 percent in an OLS estimation and 28 percent in an IV estimation, an effect that was statistically significant at the 5 percent level. The Abdulai et al. (2011) study reconfirmed previous findings for Ghana.
They found that being an owner cultivator enhanced the propensity of investment in soil- improving measures between 11 and 49 percent. Deininger and Jin’s (2006) study of
Ethiopia showed that improvement in transfer rights increased investment in tree planting by 7 percent and in building terraces by 5.5 percent. Beside the studies from Africa,
Alston et al.’s (1999) case study of Brazil’s frontier found that title increased investment in land by 29 percentage points in the Altamira region. The empirical analysis in this study indicates that the estimated coefficient on land rights is statistically significant but very small relative to the estimates from other econometric studies of particular countries.
2.6 Conclusion
This study uses propensity score matching to empirically test the effect of de jure land rights on land investment after land reform in 1989. Matching results using land title and a combined measure reveal that there are positive links between de jure land rights and investment in land. Investment on titled farms is 1.5-1.7 percentage points higher than untitled farm. When application receipts are used as the measure of lands rights, the estimated coefficient on land rights is not statistically significant at the 10 percent level.
This result indicates that land holders with application receipts do not view the rights adhering to the receipts as providing more land tenure security than a de facto claim.
Results in this study are similar to other studies conducted in Asia, Latin America, and
Africa that find a positive link between de jure land rights and agricultural outcomes.
65
However, the magnitude of the impact of land rights on land investment in this analysis is relatively small if compared to other studies of the field.
The paper also extends the discussion of issues in previous empirical studies of property rights related to measures of land investment, endogeneity of land rights, soil quality, and sample size. These issues have not been widely discussed previously and inconclusive results of studies measuring the impact of de jure land rights on investment may be related to these issues (Fenske, 2011; Deininger & Jin, 2006).
The findings in this study contribute to the empirical study on the relationship between de jure land rights and investment in low-income countries. Given the significance of investment to poverty alleviation in the long run, it is important to understand the role and the commitment of the government to implementation of its land titling programs. De jure rights were established in Cambodia but institutions registering and enforcing land rights did not function well. The failure of the central cadastral data registry to record land transfers and maintain an up-to-date record of land ownership is a huge problem (So, 2009). Moreover, informal land transfers continue (Chan et al., 2001).
These problems clearly impede the sustainability and quality of the cadastral system and pose important obstacles for the future prosperity of Cambodia’s farmers.
66
Table 2.1
Land investment: 1979-2004
Note: Each column represents a total number of land investment conducted by households on their plot from 1979 to 2004. Source: All data are from CSES (2003-2004).
67
Figure 2.1
Land investment: 1979-2004
Note: Vertical axis represents a percentage of sample farms undertaking eight types of investment. Source: All data are from CSES (2003-2004).
68
Table 2.2
Farm characteristics: Summary statistics
Means Std. Dev. Type of document Land title certificate (%) 0.21 0.41 Application receipt (%) 0.24 0.43 Land investigation document (%) 0.01 0.11 Rental document (%) 0.00 0.06 Other documents (%) 0.02 0.12
Mode of acquisition Given by state (%) 0.45 0.50 Inherited from families (%) 0.32 0.47 Bought land (%) 0.12 0.33 Occupied for free (%) 0.06 0.24 Received from friend (%) 0.00 0.02 Rented plot (%) 0.03 0.18 Others (%) 0.00 0.08
Type of plot Wet land (%) 0.67 0.47 Dry land (%) 0.11 0.31 Wet and dry land (%) 0.01 0.11 Crop land (%) 0.14 0.35 Vegetable land (%) 0.02 0.14 Idle land (%) 0.02 0.14 Others type (%) 0.03 0.16 Irrigated land in dry season (%) 0.09 0.07 Irrigated land in wet season (%) 0.19 0.26 Irrigated land in dry and wet seasons (%) 0.05 0.09
Investment on plot since acquisition Digging well and ditch (%) 0.05 0.32 Terracing (%) 0.00 0.02 Drainage construction (%) 0.05 0.07 Soil reclamation (%) 0.00 0.08 Establish orchard (%) 0.03 0.16 Perennial trees (%) 0.02 0.13 Others (%) 0.00 0.08
Others Duration of tenure (Years) 18.14 8.64 Ever litigated on plot (%) 0.02 0.13 Plot can be used as collateral (%) 0.77 0.42 Number of observations 19,764
Source: All data are from CSES (2003-2004).
69
Table 2.3
Household loans: Summary statistics
Mean Std. Dev. Purpose of loan Agricultural purpose (%) 0.24 0.43 Household consumption (%) 0.32 0.47 Household medical expense (%) 0.12 0.33
Sources of loan Relative in Cambodia (%) 0.27 0.44 Friend (%) 0.14 0.35 Money lender (%) 0.25 0.43 Bank (%) 0.036 0.18 NGOa (%) 0.16 0.37 Amount average borrowedb (Reil) 953,324 3,706,033 Monthly interest rate (%) 0.04 0.08 Number of observations 8,024 Note: a. NGO stands for non-governmental organization. b. Average amount borrowed by households in Cambodian currency. Source: All data are from CSES (2003-2004).
70
Table 2.4
Household characteristics: Summary statistics
Means Std. Dev. Education of household head (years) 4.37 4.02 Age of head (years) 45.14 13.74 Female headed (%) 0.22 0.41 Household size (number) 4.94 1.98 Male member aged 10 to 14years old (number) 0.36 0.60 Female member aged 10 to 14years old (number) 0.35 0.59 Male members aged 15 to 55 years old (number) 1.29 0.94 Female members aged 15 to 55 years old (number) 1.45 0.90 Male members aged over 55 years old (number) 0.19 0.39 Female members aged over 55 years old (number) 0.26 0.45 Value of live stock (Reil)a 1,437,520 2,688,887 Value of durable goods (Reil) 1,977,184 6,197,216 Number of rooms in home 1.36 0.78 Number of plots per household 1.31 1.24 Size of land operated (hectares) 1.29 0.45 Access to all-weather road (kilometer) 3.02 14.51 Number of observations 14,984 Note: a. Average value of live stock and durable goods of household in Cambodian currency. Source: All data are from CSES (2003-2004).
71
Panel A: Land title
Table 2.5a
Propensity score estimation result for titled farm regression
Covariates Estimated z-statistics coefficients Education of household head 0.090 7.78 Age of head 0.030 2.92 Female headed 0.055 1.06 Member age below 10 -0.088 -0.72 Male members 10-55 0.002 0.10 Female members 10-55 0.050 2.87 Value of live stocks (*1000) -0.005 -1.24 Value of durable goods (*1000) 0.002 4.22 Number of rooms in home 0.063 2.11 Size of land operated 0.007 1.78 Wet land 0.261 2.79 Crop land 0.984 6.11 Wet and dry land 0.093 1.00 Access to all-weather road 0.436 9.51 PseudoR2 0.016 Log-Likelihood -9505.31 Chi 2 300.73 Number of observations 17,166 Source: All data are from CSES (2003-2004).
72
Table 2.5b
Results from PSM: Land titles
Number of cases ATTa t-statistic
Radius matching Treatment 3,830 0.017 2.89 Control 12,987
Kernel matching Treatment 3,949 0.015 2.75 Control 14,815
Stratification matching Treatment 3,948 0.016 3.06 Control 14,816 Note: a. Average effect of treatment on treated (ATT) b. Bootstrapping applied with the results reported above.
Figure 2.2
Estimated propensity scores: Land titles
.2 .4 .6 .8 1 Propensity Score
Untitled farm Titled farm
73
Table 2.5c
Treated vs. controls (land titles)
Variables Treated Control % bias t-sta. p>/t/ mean mean Education of head Unmatched 4.351 3.952 10.8 6.000 0.000 Matched 4.350 4.281 1.8 0.81 0.420
Age of head Unmatched 47.054 44.909 15.8 8.78 0.000 Matched 47.051 46.87 1.3 0.60 0.548
Female headed Unmatched 0.180 0.179 0.1 0.04 0.971 Matched 0.180 0.187 -1.9 -0.82 0.411
Member below 10 Unmatched 1.171 1.249 -5.1 -2.83 0.005 Matched 1.171 1.196 -1.6 -0.71 0.479
Male members 10-55 Unmatched 2.023 1.935 7.4 4.12 0.000 Matched 2.023 2.007 1.4 0.61 0.544
Female members 10-55 Unmatched 2.239 2.109 11.5 6.44 0.000 Matched 2.239 2.217 2.0 0.86 0.390
Value of live stocks Unmatched 6.394 6.355 1.5 0.83 0.407 Matched 6.394 6.411 -0.6 -0.28 0.778
Value of durable goods Unmatched 6.153 5.856 17.6 9.86 0.000 Matched 6.153 6.097 3.3 1.51 0.132
Number of rooms in home Unmatched 1.333 1.271 9.8 5.71 0.000 Matched 1.332 1.305 4.4 1.88 0.06
Size of land operated Unmatched 1.966 1.693 4.3 3.11 0.002 Matched 1.858 1.790 1.1 0.68 0.496
Wet land Unmatched 0.706 0.689 3.7 2.04 0.041 Matched 0.705 0.708 -0.7 -0.30 0.765
Crop land Unmatched 0.120 0.133 -4.0 -2.22 0.026 Matched 0.120 0.118 0.5 0.22 0.826
Wet and dry land Unmatched 0.020 0.093 9.4 5.95 0.000 Matched 0.020 0.018 1.8 0.71 0.478
Access to all-weather road Unmatched 2.284 3.713 -10.6 -5.46 0.000 Matched 2.285 2.408 -0.9 -0.52 0.602
On common support Untreated 14,817 Treated 3,948
Off common support Untreated 0 Treated 1
74
Panel B: Land title and application receipt
Table 2.6a
Propensity score estimation result of a combined measure regression
Covariates Estimated z-statistics coefficients Education of household head 0.13 13.79 Age of head 0.013 10.64 Female headed 0.19 4.44 Member age below 10 -0.012 -1.22 Male members 10-55 0.001 0.08 Female members 10-55 0.058 4.12 Value of live stocks (*1000) 0.09 3.63 Value of durable goods (*1000) 0.001 6.73 Number of rooms in home 0.10 3.76 Size of land operated 0.003 0.92 Wet land 0.068 1.10 Crop land 0.52 3.48 Wet and dry land -0.078 -1.07 Access to all-weather road 0.37 10.48 PseudoR2 0.027 Log-Likelihood -12567.77 Chi 2 705.20 Number of observations 17,166
Source: All data are from CSES (2003-2004).
75
Table 2.6b
Results from PSM: Combined measure
Number of cases ATTa t-statistic
Radius matching Treatment 7,879 0.010 2.04 Control 9,165
Kernel matching Treatment 8,464 0.011 3.47 Control 10,270
Stratification matching Treatment 8,464 0.010 2.70 Control 10,270 Note: a. Average effect of treatment on treated (ATT) b. Bootstrapping applied with the results reported above.
Figure 2.3
Estimated propensity scores: Combined measure
0 .2 .4 .6 .8 1 Propensity Score
Untitled farm Titled farm
76
Table 2.6c
Treated vs. controls (combined measure)
Variables Treated Control % bias t-sta. p>/t/ mean mean Education of head Unmatched 4.343 3.783 15.1 10.30 0.000 Matched 4.343 4.378 0.6 0.41 0.678
Age of head Unmatched 46.693 44.262 17.9 12.17 0.000 Matched 46.693 46.585 0.8 0.52 0.602
Female headed Unmatched 0.185 0.174 2.8 1.91 0.056 Matched 0.185 0.186 -0.2 -0.10 0.918
Member below 10 Unmatched 1.180 1.277 -6.3 -4.27 0.000 Matched 1.180 1.190 -0.7 -0.46 0.649
Male members 10-55 Unmatched 2.004 1.912 7.7 5.28 0.000 Matched 2.004 2.001 0.3 0.20 0.845
Female members 10-55 Unmatched 2.215 2.071 12.7 8.66 0.000 Matched 2.215 2.206 0.6 0.38 0.700
Value of live stocks Unmatched 6.341 6.355 -1.6 -1.08 0.27 Matched 6.341 6.399 -2.2 -1.43 0.15
Value of durable goods Unmatched 6.331 5.758 21.2 14.48 0.000 Matched 6.331 6.085 1.7 1.13 0.257
Number of rooms in home Unmatched 1.329 1.246 13.5 9.31 0.000 Matched 1.329 1.314 2.4 1.46 0.523
Size of land operated Unmatched 1.767 1.737 0.6 0.41 0.681 Matched 1.767 1.811 -0.9 -0.57 0.569
Wet land Unmatched 0.696 0.689 1.4 0.94 0.345 Matched 0. 696 0.700 -1.0 -0.63 0.528
Crop land Unmatched 0.125 0.135 -3.0 -2.04 0.041 Matched 0.125 0.121 1.2 0.80 0.423
Wet and dry land Unmatched 0.015 0.008 6.0 4.17 0.000 Matched 0.015 0.013 1.2 0.73 0.464
Access to all-weather road Unmatched 0.794 0.713 18.8 12.75 0.000 Matched 0.794 0.795 -0.3 -0.20 0.841
On common support Untreated 10,287 Treated 8,479
Off common support Untreated 0 Treated 0
77
Panel C: Application of receipt
Table 2.7a
Propensity score estimation result of application receipt farm regression
Covariates Estimated z-statistics coefficients Education of household head 0.13 11.01 Age of head 0.021 2.24 Female headed 0.225 4.20 Member age below 10 -0.012 -1.00 Male members 10-55 -0.003 -0.16 Female members 10-55 0.042 2.41 Value of live stocks (*1000) 0.04 5.51 Value of durable goods (*1000) 0.03 5.06 Number of rooms in home 0.110 3.48 Size of land operated -0.040 -4.72 Wet land -0.064 -0.88 Crop land -0.138 -1.62 Wet and dry land 0.054 0.28 Access to all-weather road 0.37 10.48 PseudoR2 0.023 Log-Likelihood -8003.54 Chi 2 421.87 Number of observations 14,810
Source: All data are from CSES (2003-2004).
78
Table 2.7b
Results from PSM: Application receipts
Number of cases ATTa t-statistic
Radius matching Treatment 4,225 0.007 1.15 Control 8,624
Kernel matching Treatment 4,524 0.008 1.85 Control 10,286
Stratification matching Treatment 4,524 0.006 1.06 Control 10,286 Note: a. Average effect of treatment on treated (ATT) b. Bootstrapping applied with the results reported above.
Figure 2.4
Estimated propensity scores: Application receipts
0 .2 .4 .6 .8 Propensity Score
Untitled farm Application receipt farm
79
Table 2.7c
Treated vs. controls (application receipts)
Variables Treated Control % bias t-sta. p>/t/ mean mean Education of head Unmatched 4.33 3.783 14.8 8.27 0.000 Matched 4.33 4.353 -0.6 -0.31 0.755
Age of head Unmatched 46.375 44.261 15.5 8.68 0.000 Matched 46.375 46.223 1.1 0.53 0.594
Female headed Unmatched 0.190 0.175 4.1 2.33 0.020 Matched 0.190 0.196 -1.4 -0.66 0.506
Member below 10 Unmatched 1.188 1.277 -5.7 -3.19 0.001 Matched 1.188 1.182 0.4 0.21 0.836
Male members 10-55 Unmatched 1.987 1.912 6.2 3.54 0.000 Matched 1.987 1.964 1.9 0.92 0.358
Female members 10-55 Unmatched 2.191 2.070 10.6 5.98 0.000 Matched 2.191 2.181 0.8 0.40 0.693
Value of live stocks Unmatched 2478.4 2120.4 -3.4 -1.94 0.052 Matched 2478.4 2371.5 1.8 0.82 0.410
Value of durable goods Unmatched 6.075 5.757 19.1 10.69 0.000 Matched 6.075 6.055 1.2 0.58 0.561
Number of rooms in home Unmatched 6.291 6.382 12.7 7.38 0.000 Matched 6.291 6.243 2.0 0.90 0.367
Size of land operated Unmatched 1.558 1.737 -5.6 -3.17 0.002 Matched 1.558 1.652 -2.9 -1.42 0.156
Wet land Unmatched 0.688 0.689 -0.3 0.94 0.345 Matched 0. 688 0.686 0.3 1.70 0.089
Crop land Unmatched 0.129 0.136 -1.7 -2.04 0.041 Matched 0.129 0.135 -1.5 -0.59 0.554
Wet and dry land Unmatched 0.010 0.008 1.5 4.17 0.000 Matched 0.010 0.010 -0.7 -0.17 0.388
Access to all-weather road Unmatched 0.775 0.713 14.2 7.84 0.000 Matched 0.775 0.790 -3.6 -1.79 0.074
On common support Untreated 10,286 Treated 4,524
Off common support Untreated 0 Treated 0
80
References
Aakvik, A. (2001). Bounding a matching estimator: The case of a Norwegian training program. Oxford Bulletin of Economics and Statistics, 63(1), 115-143.
Abdulai, A., Owusu, V., & Goetz, R. (2011). Land tenure differences and investment in land improvement measures: Theoretical and empirical analyses. Journal of Development Economics, 96(1), 66-78.
Atwood, D. (1990). Land registration in Africa: The impact on agricultural production. World Development, 18(5), 659-671.
Alchian, A., & Demsetz, H. (1973). The property rights paradigm. Journal of Economic History, 33(1), 16-27.
Alston, L., Libecap, G., & Mueller, B. (1999). Titles, conflict, and land use. Ann Arbor: University of Michigan Press.
Becker, S., & Caliendo, M. (2007). Mhbounds-sensitivity analysis for an average treatment effects. The Stata Journal, 7(1), 71-83.
Becker, S., & Ichino, A. (2002). Estimation of average treatment effects based on propensity scores. The Stata Journal, 2(4), 358-377.
Benjamin, D. (1995). Can unobserved land quality explain the inverse productivity relationship. Journal of Development Economics, 46(1), 51-84.
Besley, T. (1995). Property rights and investment incentives: Theory and evidence from Ghana. Journal of Political Economy, 103(5), 903-937.
Binswanger, H., Deninger, K., & Feder, G. (1995). Power, distortions, revolt, and reform in agricultural land relations. In J. Behrman, & T. N. Srinivasan (Eds), The handbook of development economics (pp.2659-2764). Amsterdam: Elsevier.
Brasselle, A., Gaspart, F., & Platteau, J. P. (2002). Land tenure security and investment incentive puzzling evidence from Burkina Faso. Journal of Development Economics, 67(2), 373-418.
Bryson, A., Dorsett, R., & Purdon, S. (2002). The use of propensity score matching in the evaluation of labor market policies (Policy Studies Institute and National Centre for Social Research Working Papers No. 4). London.
Caliendo, M., & Kopeinig, S. (2008). Some practical guidance for the implementation of propensity score matching. Journal of Economic Surveys, 22(1), 31-72.
81
Carter, M., & Olinto, P. (2003). Getting institutions “right” for whom? Credit constraints and the impact of property rights on the quantity and composition of investment. American Journal of Agricultural Economics, 85(1), 173-186.
Chan, S., Tep, S., & Sarthi, A. (2001). Land tenure in Cambodia: A data update (CDRI Working Paper 19). Phnom Penh: CDRI.
Deaton, A. (1997). The analysis of household survey. Baltimore, Maryland: The Johns Hopkins University Press.
Dehejia, R., & Wahba, S. (1999). Causal effects in non-experimental studies: reevaluating the evaluation of training programs. Journal of the American Statistical Association, 94 (448), 1053-1062.
Dehejia, R., & Wahba, S. (2002). Propensity score matching methods for non- experimental causal studies. Review of Economics and Statistics, 84 (1), 151-161.
Demsetz, H. (1967). Toward a theory of property rights. American Economics Review, 57(1), 347-359.
De Sotto, H. (2000). The mystery of capital: Why capitalism triumphs in the west and fails everywhere else. New York: Basic Books.
Do, Q. T., & Iyer, L. (2008). Land titling and rural transition in Vietnam. Economic Development and Cultural Change, 56(3), 531-579.
Deininger, K., & Jin, S. (2006). Tenure security and land-related investment: Evidence from Ethiopia. European Economic Review, 50(4), 1245-1277.
Feder, G., & Feeny, D. (1991). Land Tenure and property rights: Theory and implication for development policy. The World Bank Economic Review, 5(1), 135-153.
Feder, G., & Nishio, A. (1998). The benefits of land registration and titling: Economic land social perspectives. Land Use Policy, 15(1), 25-43.
Feder, G., Onchan, T., Chanlamwong, Y., & Hongladarom, C. (1988). Land policies and farm productivity in Thailand. Baltimore: Johns Hopkins University Press.
Fenske, J. (2011). Land tenure and investment incentives: Evidence from West Africa. Journal of Development Economics, 95(2), 137-156.
Field, E. (2007). Entitled to work: Urban property rights and labor supply in Peru. The Quarterly Journal of Economics, 122 (4), 1562-1602.
Field, E., & Torero, M. (2006). Do property titles increase credit access among the urban poor? Evidence from a nationwide titling program. Mimeo, Harvard University.
82
Galiani, S., & Schargrodsky, E. (2004). Effects of titling on child health. Economics and Human Biology, 2(3), 353-372.
Galiani, S., & Schargrodsky, E. (2010). Property rights for the poor: The effects of land titling. Journal of Public Economics, 94(9-10), 700-729.
Gavian, S., & Fafchamps, M. (1996). Land tenure and allocative efficiency in Niger. American Journal of Agricultural Economics, 78(2), 460-471.
Goldstein, M., & Udry, C. (2008). The profits of power: Land rights and agricultural investment in Ghana. Journal of Political Economy, 116(6), 981-1022.
Hayes, J., Roth, M., & Zepeda, L. (1997). Productivity in Gambian agriculture: A generalized probit analysis. American Journal of Agricultural Economics, 79(2), 369-382.
Heckman, J., Ichimura, H., & Todd, P. (1997). Matching as an econometric evaluation estimator: Evidence from evaluating a job training programme. Review of Economic Studies, 64(4), 605-654.
Heckman, J., Ichimura, H., Smith, J., & Todd, P. (1998). Characterizing selection bias using experimental data. Econometrica, 66(5), 1017-1099.
Imbens, G. W., & Wooldridge, J. M. (2009). Recent developments in the econometrics of program evaluation. Journal of Economic Literature, 47(1), 5-86.
Imbens, G. W., & Rubin, D. B. (2010). Causal inference in statistics and the social sciences. Cambridge and New York: Cambridge university press.
Jacoby, H., Li, G., & Rozelle, S. (2002). Hazards of expropriation: Tenure insecurity and investment in rural China. The American Economic Review, 92(5), 1420-1447.
Jacoby, H., & Minten, B. (2007). Land titles, investment, and agricultural productivity in Madagascar. The World Bank Economic reviews, 21(3), 461-485.
Jansen, K., & Roquas, E. (1998). Modernizing insecurity: The land titling project in Honduras. Development and Change, 29 (1), 81-106.
Kenjiro, Y. (2005). Why illness causes more serious economic damage than crop failure in rural Cambodia. Development and Change, 36(4), 759-783.
Markussen, T. (2008). Property rights, productivity, and common property resources: Insights from rural Cambodia. World Development, 36(11), 2277-2296.
83
Migot-Adholla, S., Hazell, P., Blarel, B., & Place, F. (1991). Indigenous land rights systems in Sub-Saharan African: A constraint on productivity. The World Bank Economic Reviews, 5(1), 155-175.
National Institute of Statistics. (2004). CSES 2003-2005 data management manual. Phnom Penh: Cambodia
National Institute of Statistics and Ministry of Planning. (2007). General population census, 2008. Phnom Penh: Cambodia
Peikes, D., Moreno, L., & Orzol, S. (2008). Propensity score matching: A note of caution for evaluators of social programs. Statistical Practice, 62(3), 222-231.
Pinckeny, T., & Kimuyu, P. (1994). Land tenure reform in East Africa: Good, bad or unimporant. Journal of African Economies, 3(1), 1-28.
Place, F., & Hazell, P. (1993). Productivity effects of indigenous land tenure systems in Sub-Saharan African. American Journal of Agricultural Economics, 75(1), 10-19.
Place, F., & Migot-Adholla, S. (1998). The economic effects of land registration on small holder farms in Kenya: Evidence from Nyeri and Kakamega districts. Land Economics, 74(3), 360-373.
Place, F., & Otsuka, K. (2002). Land tenure systems and their impacts on agricultural investments and productivity in Uganda. Journal of Development Studies, 38 (6), 105-124.
Platteau, J. P. (1996). The evolutionary theory of land rights as applied to Sub-Saharan African: A critical assessment. Development and Change, 27(1), 29-86.
Rosenbaum, P. R. (2002). Observational Studies. New York: Springer.
Rosenbaum, P., & Rubin, D. (1983). The central role of the propensity score in observational studies for causal effects. Biometrika, 70(1), 41-55.
Rattanabirabongse, V., Nettle, K., & Burns, T. (1998). The Thailand land titling project- the thirteen years of experience. Land Use Policy, 15(1), 3-23.
Sarthi, A., & Chan, S. (2002). Farm size, productivity, and earnings. Cambodia Development Review, 6(4), 9-11.
Siaensi, B (2004). An evaluation of the active labor market programmes in Sweden. The Review of Economics and Statistics, 86(1), 133-155.
84
So, S. B. (2009). Political economy of land registration in Cambodia. (Doctoral dissertation). Retrieved from ProQuest Dissertations and Theses. (Accession Order No. AAT 3359039).
SMERU. (2002). An impact evaluation of systematic land titling under the land administration project (LAP). Jakarta: SMERU Research Team.
Smith, J., & Todd, P. (2005). Does matching overcome Lalonde’s critiques of nonexperimental estimators?. Journal of Econometrics, 125(1-2), 305-353.
Smith, R. (2004). Land tenure, fixed investment, and farm productivity: Evidence from Zambia’s Southern province. World Development, 32(10), 1641-1661.
World Bank. (2002). LMAP project appraisal documents report no. 22869-KH. Phnom Penh: World Bank.
______. (2006). Cambodia halving poverty by 2015? Poverty assessment 2006. Phnom Penh: World Bank.
Yu, B., & Diao, X. (2011). Cambodia’s agricultural strategy: Future development options for the rice sector (CDRI Policy discussion paper). Phnom Penh: CDRI.
Zhao, Z. (2000). Using matching to estimate treatment effects: data requirements, matching metrics and Monte Carlo evidence. Review of Economics and Statistics, 86(1), 91-107.
85
CHAPTER THREE
PROPERTY RIGHTS IN LAND AND HOUSEHOLD CONSUMPTION
3.1 Introduction
For some time now, the argument that de jure land rights would lead to a range of social benefits has led development practitioners to advocate for land reform programs oriented towards registration of private land rights. The potential social benefit of de jure land rights stems from better incentives for land investment, better access to credit markets, and the ability to secure land market transactions. Secured land rights allow legitimate owners of land to enjoy the fruits of investment in their land (Besley, 1995;
Alchian & Demsetz, 1973; Demsetz, 1967). Similarly, clear proof of ownership of the land allows legitimate land owners to access formal credit markets to finance investment in the land due to their ability to post land as collateral (Feder & Nishio, 1998). Further, registered de jure land rights allow for secured transfer of land from inefficient to efficient users (Jacoby & Minten, 2007).
Previous studies have posited that de jure land rights reform in developing countries had the potential to reduce poverty and enhance economic growth (World
Bank, 1975; Besley & Burgess, 2000; Deininger, 2003; Field, 2007). De jure enforcement and titling of arable lands are often posited as factors that facilitate low- income households climbing out of poverty by allowing land holdings to serve as collateral for loans to improve farm productivity and to smooth consumption of farm households (De Soto, 2000).
86
The increases in land investment, the greater access to capital markets, and the
improvement in the functioning of land rental and sales markets can create an enabling
environment for overall economic growth as well as improve the farm household’s well- being. Numerous empirical studies of the impacts of establishing de jure land rights have
found, however, that the impacts vary widely across countries. Some studies indicated a
positive link between de jure land rights and land investment (Jacoby, Li, & Rozelle,
2002; Alston, Libecap, & Mueller, 1999; Besley, 1995), as well as access to credit
markets (Field & Torero, 2006; Carter & Olinto, 2003; Feder, Onchan, Chanlamwong, &
Hongladarom, 1988) while others did not (Jacoby & Minten, 2007; Migot-Adholla,
Hazell, Blarel, & Place, 1991). Over the last decade, the literature studying the impact of
de jure land property rights has greatly expanded its scope, estimating the effect of de jure land rights on labor supply, child health, housing, and human capital investments
(Galiani & Schargrodsky, 2010; Field, 2007; Galiani & Schargrodsky, 2004). These
studies have typically provided evidence in support of the theory of de jure property
rights in land. Other studies indicated that land reform and increased access to land have
positive and statistically significant impacts on poverty reduction and welfare for a broad
spectrum of farm households (Besley & Burgess, 2000; Finan, Sadoulet, & de Janvry,
2005).
This Chapter examines the relationship between de jure land rights and household
consumption by using data from the 2003-2004 Cambodia Socio-Economic Survey
(CSES) to estimate this relationship. The survey data were collected during a period when the country’s economy was growing relatively smoothly. From 2000 to 2005, the country did not experience a major natural disaster, drought, large economic fluctuation,
87
or political upheaval, all of which could have dramatically affected household
livelihoods. The Chapter’s empirical analysis uses three different measures of de jure
land rights—the households that possess land plots with land titles, with land titles or application receipts, and just application receipts—to estimate the effect. To account for endogeneity of de jure land rights, the empirical analysis uses instrumental variable estimation techniques.
The aim of the study is to provide new evidence on the impact of formalizing land rights on household consumption in low-income countries, a topic that has not been widely studied yet. The findings should also contribute to the broader empirical literature on de jure land rights in Cambodia. Most studies of de jure land rights in Cambodia have examined various issues pertaining to the reform process including problems related to capacity of land institutions, conflicts over land rights, and land registration. Most are qualitative in nature. This study estimates the impact of formalizing land rights on a household’s consumption using one regression model that does not account for endogeneity of land rights and another that does. I find that the household’s choice to obtain de jure land rights results in a statistically significant increase in household consumption expenditures in regression estimates that do not account for endogeneity of de jure rights. When the regression includes instruments for de jure land rights, the estimated impact on household consumption becomes even larger.
3.2 Land reform and poverty reduction: A literature review
Access to land is the main source of farmers’ livelihood and their assurance of household well-being. Secured access to land provides a means of subsistence through
88
crop production and access to credit for land investment. In many agrarian economies
that experience a transition to a non-farm based economy, insecure access to land poses a major challenge to rural livelihoods and increases the severity of poverty among the rural
population. Thus, ensuring that people have secure access to land is a key to poverty
reduction, improvement of household well-being, and enhancement of economic growth
(Deininger, 2003).
The importance of land tenure security has made land policy reform one of the
key policy issues in many developing countries. Land reform encompasses a broad array
of policy issues and is not an easy task. Large-scale land reforms are often entangled with political uprisings that stem, in part, from rising landlessness and inequality in land holdings. In this context, land reform programs aim to address food security and avoid social unrest by making land available to the landless and land poor families. Such reforms imply the redistribution of land from landlords and land speculators to those who physically cultivate it and to those who experience inadequate access to food
(Barraclough, 1991).
The content of land reform varies across countries and is driven by specific local needs, although the basic idea of land reform is common to all countries. Land reform in
Latin America has meant redistribution of land resources to rural populations or changes in the size of landholding whereas in many parts of Asia, land reform has meant making land available to the tillers (Maxwell & Wiebe, 1999). In many of the post-communist states, including Cambodia, land reform entails two stages: (1) the privatization of land holdings and (2) the improvement of the institutional mechanism that protects and registers private property rights in land. The reform that involves privatization of land
89
often occurred in the context of withdrawing communist ideology at the end of the cold
war and was designed to increase farmers’ sense of ownership, promote productivity on
land, and increase conservation. Land privatization requires that a mechanism to govern
private property rights be established so as to ensure security of private land holdings and
to support the land market. Land tenure security reform is typically associated with this latter stage of reform: land registration and titling (So, 2009). Thus, the examination of the relationship between land reform and social well-being involves the examination of how a land registration and titling program affects tenure security, thereby generating other benefits detailed by proponents of private property rights, including increases in land investment, better access to credit markets, and secured land market transactions.
Empirical literature on the relationship between land registration and social well- being can be categorized into two groups. The first group provides empirical testing of
the theory that tenure security leads to increasing long-term land investment, better access to credit, and secure land market transactions. Results from different countries are mixed. Studies from Thailand and Brazil indicate a positive relationship between land
tenure security and investment in land (Alston et el., 1999; Feder et al., 1988). Some
studies further emphasize that land title facilitates household access to credit markets
(Field & Torero, 2006; Feder et al., 1988). Findings from Ghana and Ethiopia provide
similar results: tenure security increases land investment (Goldstein & Udry, 2008;
Deininger & Jin, 2006; Besley 1995). Studies in Sub-Sahara Africa do not, however,
support these hypotheses (Jacoby & Minten, 2007; Braselle, Gaspart, & Platteau, 2002;
Place & Migot-Adholla, 1998; Place & Hazell, 1993; Migot-Adholla et al., 1991).
Further, a case study of Paraguay’s land titling program indicated that small farm holders
90
still had difficulty accessing formal credit markets after the reform was implemented
(Carter & Olinto, 2003).
The second group of literature extends the empirical testing of the implications of
the theory to include the indirect impacts that land registration may have on other
dimensions of social well-being. Various findings indicate that land reform programs
geared towards increasing security of land tenure have been associated with increases in
household labor supply, housing improvements, increases in child educational attainment
and child health, and declines in household size (Galiani & Schargrodsky, 2010; Field,
2007; Galiani & Schargrodsky, 2004). Other studies found that land reform programs reduced poverty and enhanced household welfare (Besley & Burgess, 2000; Finan et al.,
2004). This study further expands on the second group of literature by examining whether establishment of de jure land rights in Cambodia led to an increase in social well-being as measured by consumption of individual households.
3.3 Land reform and poverty reduction in Cambodia
The agenda for land reform in Cambodia is listed in the National Strategic
Development Plan (NSDP) as a tool to reduce poverty and enhance economic growth
(NSDP, 2009). With a series of land reforms, most households were expected to enjoy such benefits as increases in land tenure security, access to credit markets, greater land investment, and increased agricultural productivity. Poor and vulnerable groups were expected to gain the most from the reforms (World Bank, 2002). At a glance, land reform in Cambodia is in line with several elements stated in the theory as well as successful practices of land titling programs in neighboring countries. However, some researchers
91
have expressed concerns over the social benefits of de jure land rights because the
number of land disputes rose after the reform started and was still in progress, although this may have been due in part to an increase in land value and economic expansion
(Chan et al., 2001). Between 1992 and 1993, the United Nations Transitional Authority in
Cambodia (UNTAC) received many complaints related to boundary issues and land grabs by civilians and soldiers (So, 2009; Ledgerwood, 1998). The appeals court received 1,325 land dispute cases between 1995 to mid-1999; in 2001 alone, the court received 1,310 cases (Cooper, 2002). The data from CSES 2004 indicated that approximately 2% of
cultivated plots were litigated plots.
Other challenges to land reform in Cambodia are the capacity of land governance
institutions, infrastructure in rural areas, and access to formal credit market. Although the
land reform in 2002 aimed to strengthen institutional capacity so that the full benefit of
land titling could be realized, land governance institutions continued to perform poorly.
The complex bureaucratic procedure required to register land ownership transfers and the
corruption associated with this registration process have pushed households away from
the formal procedures for transferring registered de jure land titles. Instead, households have continued to use traditional processes for transferring land, as the process is easier and entails lower cost to the parties. As a result, land ownership records quickly are becoming outdated, threatening the overall effort to build a centralized land registry through the SLR and reducing the value of a de jure title (So, 2009; Chan et al., 2001).
Poor infrastructure in rural areas is another bottleneck to the improvement of living conditions of rural households, as it prevents farmers from distributing their products to market on time. Access to market information is very limited. The average
92
distance to an all-weather road is approximately 2.64 km, a taxi-stop 5.42 km, and a bus- stop 17.30 km (see table 3.1). With only a low-quality dirt road available to most farmers, these considerable distances reduce the benefit from land titling.1 Furthermore, access to
formal credit markets by farmers with titled lands has not yet lived up to expectations.
Table 3.2 indicates that households with land titles and outstanding loans tend to borrow
from informal sources such as friends, relatives, and money lenders rather than
commercial banks. The complicated procedures imposed by commercial banks together
with a limited knowledge of the formal credit system by many households discourage
them from obtaining loans from formal credit institutions. Among households who
borrowed from formal credit institutions, only about 24 percent used their loans for
agricultural purposes. The remainder used loans for other purposes such as household
consumption (32%) and medical emergencies (12%). Several qualitative studies indicate
that expenditures for medical emergencies and household consumption often are
associated with a household’s sale of its lands and subsequently becoming landless.
Lands have been sold at fire sale prices to finance emergency needs (Kenjiro, 2005;
Damme, Leemput, Por, Hardeman, & Meessen, 2004; Huy, Wichmann, Beatty, Ngan,
Duong, Margolis, & Vong, 2009).
In spite of the several challenges discussed above, some bright sides of the reform
can be noted. A recent empirical study indicates that a de jure title for farm land in
Cambodia is positively correlated with a farm’s productivity and value (Markussen,
2008). Despite findings from various qualitative studies that land reform in Cambodia yielded mixed results, Cambodia’s poverty rate continued to fall after the program was
1 After the civil war, infrastructure was heavily damaged throughout the country. New infrastructure investment has, however, been concentrated in urban areas.
93
implemented, declining from 47% in 1994 to 35% in 2004 and to 30% in 2007 (World
Bank, 2008). Further, GDP growth also increased during the land reform period, rising
from an average annual rate of 7% over the 1994 to 2002 period (during the
implementation of the land reform program) to 8% over the 2003 to 2009 period after
implementation of the land reform programs.2 Of course, there were obviously many
other factors besides land tenure security that caused Cambodia’s GDP to increase and its
poverty rate to decline after land reforms were implemented. This Chapter’s main goal is to provide a quantitative estimate of the extent to which the property rights reform affected household well-being in Cambodia at the time of the 2004 CSES survey, roughly fourteen years after implementation of land reform began in 1989.
3.4 Empirical analysis and econometric issues
3.4.1 Empirical model
The empirical model utilized in this Chapter is based on a theoretical model
developed by Finan et al. (2004) which I modify to incorporate land tenure security into
their production function to fit the particular focus of this study. Finan et al. (2004)
derived a specification of the household income equation from a standard agrarian
household production model which yields an estimate of the effect of the size of
landholding on household income. They examined the economic return to land in the
context of multiple market imperfections. In their model, a household maximizes income
by choosing time allocation and purchased inputs:
max , , , + ( ) + ( ( )) , , 𝑓𝑓 𝑜𝑜 𝑜𝑜 𝑙𝑙𝑓𝑓 𝑙𝑙𝑜𝑜 𝐼𝐼 𝑝𝑝𝑝𝑝�𝑙𝑙 𝐼𝐼 𝑅𝑅 𝐻𝐻� − 𝑞𝑞𝑞𝑞 𝑤𝑤𝑤𝑤 𝑙𝑙 𝑟𝑟 𝑞𝑞𝑞𝑞 − 𝐾𝐾 − 𝑤𝑤𝑤𝑤 𝑙𝑙
2 Table 1.1 in Chapter One contains information on GDP growth rates from 1994 to 2009. 94
s.t.
+ , ( 0, 0)
𝑜𝑜 𝑓𝑓 𝑜𝑜 𝑜𝑜 𝑙𝑙 +𝑙𝑙 ≥ (𝐿𝐿 )𝑙𝑙+≥ ( 𝑙𝑙) ≥ (1)
𝑜𝑜 where 𝐾𝐾, , 𝑤𝑤𝑤𝑤and L𝑙𝑙 denote𝜃𝜃 𝑅𝑅 on-≥farm𝐼𝐼𝐼𝐼 labor, off-farm labor, and total hours of labor per year;
𝑓𝑓 𝑜𝑜 ( ) is𝑙𝑙 the𝑙𝑙 number of days worked as a function of off-farm labor supplied ( > 0 and
𝑜𝑜 𝑔𝑔"𝑙𝑙< 0); I, q and p denote unit of input, market price of input, and market price𝑔𝑔 of′ output;
𝑔𝑔 , , , is the production function where H represents the set of household
𝑓𝑓 characteristics𝑓𝑓�𝑙𝑙 𝐼𝐼 𝑅𝑅 𝐻𝐻� that affects the return on productive assets; R represents the amount of land cultivated; K is the initial capital; ( ) is “the capital available at an interest rate r to
a household with land endowment R”; 𝜃𝜃and𝑅𝑅 Y denotes household income.
Forming the LaGrangian, the first-order conditions are:
= (1 + + )
𝐿𝐿 𝑝𝑝𝑓𝑓 = 𝑤𝑤(1 + 𝑟𝑟 + 𝜆𝜆) 𝑔𝑔′
𝐼𝐼 𝑝𝑝[𝑓𝑓 +𝑤𝑤 ( )𝑟𝑟+ 𝜆𝜆( ) ] = 0 (2)
𝑜𝑜 Let the𝜆𝜆 superscript𝐾𝐾 𝑤𝑤𝑤𝑤 𝑙𝑙 (*) 𝜃𝜃indicate𝑅𝑅 − 𝐼𝐼𝐼𝐼the value of the choice variable that maximizes the household’s income, i.e., is the solution to the household’s income maximization problem above. Then the income equation can be written as:
= , , , (1 + ) + (1 + ) ( 0) + ∗ ∗ ∗ ∗ 𝑓𝑓 𝑌𝑌 = 𝑝𝑝𝑝𝑝( �,𝑙𝑙 ,𝐼𝐼 , 𝑅𝑅, 𝐻𝐻,� −, 𝑞𝑞,𝐼𝐼 ) 𝑟𝑟 𝑤𝑤 𝑟𝑟 𝑔𝑔 𝑙𝑙 𝑟𝑟𝑟𝑟
𝜓𝜓 𝑝𝑝 𝑞𝑞 𝑟𝑟 𝑤𝑤 𝐿𝐿 𝐾𝐾 𝑅𝑅 𝐻𝐻 = + [( (1 + ) ] + [ (1 + )] ∗ ∗ 𝑓𝑓 𝑑𝑑𝑑𝑑 ′ 𝑑𝑑𝑙𝑙 𝑑𝑑𝐼𝐼 𝑑𝑑𝑑𝑑 𝑝𝑝𝑓𝑓𝑅𝑅 𝑝𝑝𝑓𝑓𝐿𝐿 − 𝑤𝑤 𝑟𝑟 𝑔𝑔 𝑑𝑑𝑑𝑑 𝑝𝑝𝑓𝑓𝐼𝐼 − 𝑞𝑞 𝑟𝑟 𝑑𝑑𝑑𝑑 = + + + ∗ ∗ (3) 𝑓𝑓 ′ 𝑑𝑑𝑙𝑙 𝑑𝑑𝐼𝐼 𝑝𝑝𝑓𝑓𝑅𝑅 𝜆𝜆𝑔𝑔 𝑑𝑑𝑑𝑑 𝑝𝑝𝑓𝑓𝐼𝐼 𝑞𝑞𝑞𝑞 𝑑𝑑𝑑𝑑
95
The income equation generated from this household production model is a
function of input and output prices, the household’s endowment of productive assets, and
other parameters that have an impact on the return to assets. Finan et al. (2004) discussed
two cases for analyzing the effect of the household’s land holdings on its income. In the
first case, capital markets are assumed to be perfect and “the marginal value of land is
simply the value of its marginal product”. In the second case, both labor and credit
markets are assumed to be imperfect. As a result, increases in the size of land holdings
have both a direct and an indirect impact on household income. More land increases
production, gives additional flexibility in the allocation of production inputs, increases
labor allocated to on-farm agriculture work, and decreases the household’s
unemployment rate. As the size of the household’s land holding increases, “the credit
constraint may either tighten or loosen depending on the relative increase in credit
availability and on the demand for inputs”. In addition, an increase in household landholding in a region with thin labor markets will increase the marginal value of household labor and reduce the gap between the household’s shadow wage and the market wage (Finan et al., 2004, p. 31).
The household production model in Finan et al. indicates that any factor that affects the return to the productive assets of the household will affect its income. Their theoretical model focuses on the link between landholding and income and implicitly assumes that land tenure is fully secure. Lóapez and Valdés (2000) posited that poor famers in Latin America hold approximately 1-5 hectares of land and noted even a holding of five hectares might be insufficient to allow Latin American farmers to consume above the poverty line without other sources of income if their land tenure was
96
not fully secure. Lóapez and Valdés noted that most of the poorest farmers in the Latin
American countries they studied did not have title to their lands. As a consequence, farmers cannot use their land as collateral to finance farm investment. Without land title, the market value of the farmers’ lands will be lower, and this hampers the ability of households to exit poverty (Lóapez & Valdés, 2000). Cambodia also faces these particular issues. In the Cambodian context, land tenure security is one factor that affects household income. If a farm household’s land is confiscated, it will lose the portion of its income generated from this factor.
I incorporate a land tenure security variable into the household production function described above, to understand how it affects household income. Let (0,1) denote the probability that land can be seized by either the government or a private𝜏𝜏 ∈ party and S is a variable measuring land tenure security; the probability that land is not being expropriated is an increasing function of land tenure security; the probability of expropriation function is ( ); it is assumed that ( ) > 0. The household income ′ equation (3) can be rewritten𝜏𝜏 𝑆𝑆as: 𝜏𝜏 𝑆𝑆
= , , , , ( ) (1 + ) + (1 + ) ( 0) + ∗ ∗ ∗ ∗ 𝑓𝑓 𝑌𝑌 = 𝑝𝑝𝑝𝑝( �,𝑙𝑙 ,𝐼𝐼 , 𝑅𝑅, 𝐻𝐻, 𝜏𝜏, 𝑆𝑆 ,� −, 𝑞𝑞(𝐼𝐼)) 𝑟𝑟 𝑤𝑤 𝑟𝑟 𝑔𝑔 𝑙𝑙 𝑟𝑟𝑟𝑟
=𝜓𝜓 𝑝𝑝 𝑞𝑞(𝑟𝑟)𝑤𝑤>𝐿𝐿0 𝐾𝐾 𝑅𝑅 𝐻𝐻 (4)𝜏𝜏 𝑆𝑆 𝑑𝑑𝑑𝑑 ′ 𝑑𝑑𝑑𝑑 𝑆𝑆 When ( ) enters𝑝𝑝𝑓𝑓 𝜏𝜏 the𝑆𝑆 production function, the income equation is a function of input and
output 𝜏𝜏prices,𝑆𝑆 the household’s endowment of productive assets, i.e., the size of land
holding and land tenure security, and other parameters that have an impact on the return
to assets. Equation (4) implies that household income is positively associated with land
tenure security.
97
As its measure of household well-being, this Chapter uses household consumption
rather than household income, both of which are frequently used to measure this concept
in empirical studies. I use consumption because it is a more reliable measure in
comparison to income in a society where the majority of households work in agriculture
or are self-employed. Whereas household income suffers from large measurement errors
and is subject to large annual fluctuations, consumption is more reliable because
households adjust their borrowing, saving, and mutual insurance to smooth consumption
over time (Ravallion, 1996; Deaton, 1997). Additionally, household consumption data
sets in developing countries are less subject to measurement error than income, as
households may hide information on their real income to avoid taxes (Deaton, 1997).
3.4.2 Regression Estimates
The model discussed above suggests a regression of household i consumption on
land rights, household i characteristics and demographics, and household i landholding:
= + 1 + 2 + 3 + (5)
𝑖𝑖𝑖𝑖 𝑣𝑣 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 where 𝐶𝐶 is 𝛼𝛼a measure𝛼𝛼 𝑇𝑇 of 𝛼𝛼household𝑍𝑍 𝛼𝛼 𝐿𝐿i consumption𝜀𝜀 in village v, is a village fixed
𝑖𝑖𝑖𝑖 𝑣𝑣 effect, 𝐶𝐶 denotes household i who possesses land with land title 𝛼𝛼in village v, is a
𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 vector of𝑇𝑇 household i characteristics and demographics, denotes size of land operated𝑍𝑍
𝑖𝑖𝑖𝑖 by each household, and is an error term. The inclusion𝐿𝐿 of village fixed effects captures
𝑖𝑖𝑖𝑖 village-specific characteristics𝜀𝜀 , such as infrastructure, economic situation, variation of land rights across villages, and weather conditions.
98
3.4.3 Data
The paper’s sample is drawn from a sample collected by the Cambodia Socio-
Economic Survey (CSES) conducted in 2003-2004 by the National Institute of Statistics
(NIS) of the Ministry of Planning (MOP). The goal of CSES is to collect information on various socio-economic aspects of households to track down the extent of household poverty. It is a nationally representative survey of 15,000 households in 24 provinces and
900 villages conducted from November 2003 to January 2005. The survey is available at individual, household, and village levels. The individual and household datasets contain information on employment, education, consumption and income, child health, housing construction, and farmland. The village dataset contains general information on population, agricultural land, development projects, schooling, presence of a hospital, common property resources, and occurrence of natural disasters. Since Chapter Two provides a full discussion of farm characteristics in the CSES, I focus here on household characteristics and household consumption expenditures.
3.4.4 Household consumption
The CSES 2004’s consumption data are collected from two sources: monthly diaries of household expenditure and income and recall questions. These methods are commonly used in household expenditure surveys in developed and developing countries.
However, the recall data collection method carries more measurement error in comparison to the diary data collection method (Gibson, 2002).
Consumption expenditure in the CSES 2003-2004 covers 14,984 households of which 12,000 were interviewed during calendar year 2004. The consumption expenditure
99
data can be classified into three main categories. The first category contains information
on food expenditure. Households were instructed to fill out monthly diaries of household
expenditure and income that listed all of last week’s expenditure on any type of food
purchased, produced, or received in kind.3 After completion, the interviewers went
through monthly diaries with each household and filled in missing entries. The food
expenditure category also covered expenditure on tobacco and alcohol consumptions.
The second category contains detailed expenditures on monthly housing rent, utilities,
maintenance, and furniture. The third category has expenditures on personal care and
effects, clothing, recreation, medical care, education, and transportation. The reference
period for this category varies according to the type of question. The empirical analysis in
this Chapter uses a measure of household consumption expenditure constructed by the
National Institute of Statistics (NIS) from the 2003-2004 CSES. The method used to
construct Cambodian household consumption expenditure by NIS is similar to the
method illustrated in Deaton and Zaidi (2003).
Table 3.3 displays summary statistics for household characteristics such as gender
of the household head, household size and migration, and consumption expenditure. The
average education of a household head is just the fifth grade. This implies that most household heads can only read and write simple messages and have very limited general knowledge. Some may not have completed primary school.
A household has an average size of five members; most have male heads; and
about 0.09 family members were away from home for more than two months during the
five years prior to the survey date. The percentage of household heads who are self-
3 The NIS used current market prices to estimate the value of food received in kind or self produced by Cambodia households. 100
employed (71 percent) is much higher than the percentage who are paid employees (19
percent). The higher percentage of self-employed households can be explained by the fact that farmers, especially in the rural areas, cultivate their own farms.
The average land holding of each CSES household is 1.29 hectares held over 1.31 plots. The average distance from a household in each village to an all-weather road is approximately 3 kilometers. The data further show that 29 percent of households have no cultivated land. This is a relatively high figure relative to other poor developing countries. Sixteen percent of Cambodian households report that they have experienced food shortages. Average household consumption is about US$129 per month, with food expenditure accounting for more than 50 percent of the total.
3.4.5 OLS Regression Results
The discussion in this section focuses on results from Ordinary Least Squares
(OLS) estimates of equation (5) in which de jure land rights enter directly as a binary variable. I use total household consumption to explore this relationship instead of per capita household consumption. Per capita household consumption is a better measure of a household’s living standard than total household consumption, but construction of the per capita variable requires adjustments for costs of raising children, household economies of scale and gender differences which are beyond the scope of this study. To partly account for these effects, the number of children and the number of adults in the household are used as separate control variables in all regressions.
This study uses household consumption as a measure of household well-being. Of course, actual household expenditure may differ from observed household expenditure
101
because households fail to account properly for their expenditures, do not remember details of many transactions, and forget that they have made some transactions. Even if the dependent variable in the regression is measured with error, Wooldridge (2006)
demonstrated that OLS estimation would still be appropriate if the measurement error is a
random reporting error that is uncorrelated with the regression’s independent variables.
For purposes of this analysis, the main implication of the presence of measurement error
in Cambodian household consumption expenditures—this study’s dependent variable—is
that OLS estimation produces larger estimated variances of the estimated OLS
coefficients (p. 319).
This Chapter’s empirical analysis uses three measures of de jure land rights:
households who possess land with titles, with application receipts, and with either an
application receipt or a land title. Although a household with an application receipt would
need to complete several additional costly steps to obtain a land title, many households
have treated an application receipt as a land title and have mistakenly reported in the
CSES survey that they have a title for their land. This misunderstanding is common due
to the limited knowledge by many households of land institutions or their expectation that
holding application receipts provides them with some limited property rights in their land
and that this may be a better choice than holding land title in some circumstances (So,
2009). In fact, some micro finance institutions also treat application receipts as a form of
property rights, often accepting them as collateral. An initial appraisal is that the tenure
security given under an application receipt is less than the tenure security provided by de
jure land rights (i.e., a title) and more than the tenure security provided by de facto rights.
Given the potentially different household expectations of the land tenure security
102
provided by a title, an application receipt, and a de facto right, separate regressions are estimated for each measure of land rights to test whether they have differential impacts on household consumption.
Table 3.4 displays results for OLS estimates using land title as a measure of land rights. The regression specifications include other factors influencing household consumption, such as household characteristics, size of land operated, distance to all- weather road, type of land, and village dummy variables. The results are broadly consistent with the predictions of economic models of land tenure security: the estimated coefficient on land rights has a positive effect on household consumption. Households with titled parcels consume 2.7 percent more than households with untitled parcels and the effect is statistically significant at the 5 percent level.4
Estimated coefficients for all control variables have the expected sign. An additional year of education for the household head raises household consumption by 3 percent and is statistically significant at the 1 percent level. Household consumption
increases with the age of the household head and decreases with age squared, with the
maximum level reached at 95 years. Female-headed households are estimated to have
lower consumption expenditures than male-headed ones. Consumption by an additional
female member age 0 to 4 and 10 to 14 is larger than consumption for an additional male
member but the result is reversed for adults ages 15 to 54. The estimated coefficient on
self-employment is negative and not statistically significant at ten percent level while the
estimated coefficient on paid employment is also negative but statistically significant at
the 10 percent level.
4 I use Kennedy’s method to calculate the effect of land rights on household consumption since land rights are dummy variables and household consumption is in logs. See Garderen and Shah (2002) for details. 103
The estimates discussed above also control for household migration. All
regressions include a variable measuring the number of household members who have been away from home for more than two months in the past five years. Finan et al. (2004) suggested controlling for household migration to account for any potential bias in the
measurement of the household labor force, as members of households with smaller farms
are likely to migrate more. In Cambodia, migration of a member of a household could have a positive impact on its total consumption due to remittances from members who have migrated to foreign countries or urban areas within Cambodia. Remittances could
help smooth household consumption and also provide an opportunity to finance expenses
associated with obtaining a land title. Finan et al. (2004) found that the estimated
coefficients on domestic and foreign migration both have positive and statistically
significant effects on household welfare.
I find the opposite result: The estimated coefficient on “Number of Household
Members Away From Home” is negative and statistically significant at the 5 percent
level. This result implies that total household consumption expenditures fall as an
additional member temporarily migrates. This may indicate the fall in household income
resulting from the use of less labor in production is not offset by remittances from
members who have migrated.
The estimated coefficient on the size of operated land is positive, increasing
household consumption by 1.2 percent for each additional hectare of land, and is
statistically significant at the 1 percent level. This result is in line with results reported by
Finan et al. (2004) who found that for smaller land holders, household welfare increases
by 1.3 times the earnings of an agricultural worker with an additional hectare of land. The
104 estimated coefficient on distance from village to all-weather road is negative and statistically significant at the 1 percent level. An increase in the distance from the village to an all-weather road of one kilometer decreases household consumption by 1 percent.
When land title and application receipt are combined, the estimated coefficient on the combined land rights variable remains positive and is statistically significant at the 1 percent level (Table 3.5). The magnitude of the estimated impact is larger when the regression is estimated with the combined land rights measure. Consumption of a household with land title or application receipt is 4.7 percent higher than consumption of an untitled household, whereas the consumption of households with application receipts increases by almost the same amount, 5 percent (Table 3.6). The signs and statistical significance for estimated coefficients on all control variables remain the same when the regression is estimated with the combined rights variable.
3.4.6 Robustness of the OLS regression results
To consistently estimate 1 by OLS in equation (5) below, it requires that the regression’s error term has a mean𝛼𝛼 of zero and is uncorrelated with the regressors:
= + 1 + 2 + 3 + (5)
𝑖𝑖𝑖𝑖 𝑣𝑣 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 OLS produces𝐶𝐶 𝛼𝛼 inconsistent𝛼𝛼 𝑇𝑇 𝛼𝛼estimates𝑍𝑍 𝛼𝛼 when𝐿𝐿 there𝜀𝜀 is measurement error, omitted variable bias, or reverse causation issues. Although the results from OLS regressions show that the estimated coefficient on land rights has a positive and statistically significant impact on household consumption for all three measures of land rights, the estimates may not be consistent if the de jure land rights are endogenous. The empirical literature on land titling in developing countries shows that households with more education, wealth, and
105
political connections are more likely to participate in land titling programs. The endogeneity of land rights is accepted in the economics and political literature on titling as a critical issue and authors have used instrumental variable technique to account for selection effects in household participation in programs to convert de facto land claims
into formal de jure holdings. Besley (1995), Braselle et al. (2002), and Markussen (2008)
used the mode of land acquisition as an instrument for land rights. This study uses two
instruments: the standard mode of land acquisition variable and a variable measuring the
length of land occupation. Fenske (2011) argued that the duration of tenure may not be
excludable and that the reduced form of the land rights regression is likely to suffer from
omitted variables bias. He argued that famers would make an effort to protect the plots
that are suitable for investment which would lead them to hold such plots longer and
make more investment. Outcomes may be “due to the unobserved plot characteristics”
(p.140). For this reason, I include a control variable for the type of land in the regression
to partially account for plot characteristics.
Wooldridge (2002) posits that a good instrument has to satisfy two conditions to
identify coefficient estimates. First, the instrumental variable must be correlated with the
endogenous variable. Second, the instrumental variables should not be correlated with the
error term. The first condition can be tested via a reduced form equation for the
endogenous explanatory variable. The second condition in most cases cannot be tested as
it involves unobserved variables (p. 83).
The results from a reduced form regression equation for land rights (see Table
3.7) show that the estimated coefficients on mode of land acquisition and duration of tenure are statistically significant. The land title and application receipt variables are
106
positively associated with the acquisition of land from the government, inheritance, and purchase, whereas gifted land and cleared land are negatively associated with land rights.
Further, the land title and application receipt variables are positively related to duration of tenure. An additional year of occupation of a claimed plot of land increases the
probability of the household obtaining land title by 0.2 percent and the effect is
statistically significant at the 1 percent level. The F-test on the joint significance of the instruments (reported in table 3.7) under the hypothesis that they are jointly zero is rejected.
In addition to the results from the reduced form regression estimation, I make a
conceptual argument that the two instruments should have a positive impact on the
probability that a household will choose to apply for de jure rights. First, households who
obtained land from the government or purchased it have a higher chance of obtaining de
jure title. As mentioned in Chapter One, starting from 1989, land in Cambodia was
privatized. The government allocated the solidarity group land to households according to
family size and availability of land in each village. The land titling program was part of
the 1989 land reform in which de facto land owners were eligible to apply for de jure
titles for their residential land and possession rights for cultivated land. Households who
obtained land through this allocation might receive an official document such as an
application receipt that could serve as a proof of ownership in a future dispute. Further,
households who purchased land from their neighbors or outsiders also tended to have
some types of document such as a sales agreement between buyer and seller witnessed at
the village or commune offices as a proof of land transaction. In general a buyer only
wants to purchase a secure land plot. Therefore, a seller has to show some type of
107
document as proof of land ownership. By contrast, households who acquired plots by clearing empty land or receiving it from friends are less likely to apply for de jure rights, at least in part because of the more tenuous nature of their de facto claims. Those households usually had no supporting documents for their occupation. Thus, the probability of obtaining a de jure title is likely to be higher for those who obtained land from the government or purchased it than those who obtained land from their friends or cleared empty public land.
Second, the duration of tenure should also increase the probability of obtaining land title, as it increases the legitimacy of the household’s claim to the land. The 2001 land law allows people who occupy land peacefully for five years to claim de jure ownership rights to their land.
Condition two requires that the mode of land acquisition and the length of land occupation are uncorrelated with the regression’s error term. Either instrument might not satisfy this condition if it should have been included in the regression equation in the first place or it is correlated with omitted variables that are relevant to the analysis but for which no available samples exist. The situation is bothersome if the correlation with omitted variables leads to a bias in the IV estimates that is much larger than the bias in the OLS estimates (Angrist & Krueger, 2001). Thus, condition two requires that the two instruments not directly affect household consumption. Regardless of how a household received its lands, their mode of acquisition does not directly affect the household’s production function or its incentive to use lands as collateral or make investments in land.
The same line of reasoning applies to the duration of tenure variable: Increased duration of tenure improves a household legal’s claim on its land but should not have a direct
108
connection with household consumption. As mentioned above, the two instruments
mainly affect land rights.
Column 3 of Tables 3.4, 3.5, and 3.6 presents the results from instrumental
variable regressions using the three measures of de jure land rights. The signs and
statistical significance of estimated coefficients for virtually all control variables are the
same as in the OLS estimates, except for the estimated coefficient on the land rights
variable. When de jure land rights are instrumented with the “mode of land acquisition” and “length of occupation” variables, the magnitude of the estimated coefficient on land rights increases from 2.7 percent to about 58 percent for the titles measure, from 4.7 percent to 20 percent for the combined measure, and 5 percent to 28 percent for the application receipts measure.5 Each of the three estimated coefficients is statistically
significant at the 1 percent level. Thus, adjusting for the endogeneity of land rights leads
to big increases in the magnitude of the estimated effect of de jure land rights on
household consumption when compared with results from OLS regressions.
I also test the overidentifying restrictions. As suggested in Wooldridge (2006), the
condition that the instrumental variables not be correlated with the error term of the
structural equation cannot be tested because it involves an unobserved error. However,
researchers can test whether some of the instrumental variables are uncorrelated with the
error term when more than one instrument is used. The result of this test for
overidentifying restrictions is reported in the bottom of each table. The test’s p-value
5 I use Kennedy’s method to calculate the effect of land rights on household consumption since land rights are binary variables and household consumption is in logs. 109
indicates that the instruments pass the overidentifying restriction test and should be
suitable instruments for the land rights variable in this analysis.6
For the three measures of land rights, the estimated coefficients on the three land
rights variables in the IV estimates are larger than those in the OLS estimates. One might
expect the opposite result, that OLS would produce an upward bias in the estimated
coefficient. However, many studies of the impact of property rights in land on households have found larger impacts in the IV estimations. For example, Besley’s (1995) study of
the link between land rights and land investment found that the OLS estimate of the
impact of rights with approval was 2.5 percent. By contrast, the IV estimate of rights with
approval was 28 percent. Goldstein and Udry (2008) and Markussen (2008) also found larger impacts in their IV estimations of the impact of property rights on household decisions. Markussen (2008) mentioned that the larger impacts in the IV estimations could signal the importance of measurement error and noted that the variable used in his study to measure land rights may not have fully captured the dimensions of de jure land
rights—a complex amalgam—that were most relevant to households and other market
participants. A larger impact in the IV estimation would be expected if the endogenous variable is binary and instruments are correlated with measurement error of the dependent variable. In this case, the IV estimations should be biased upwards (Markussen, 2008;
Kane, Rouse, & Staiger, 1999). Thus, in the presence of such measurement errors, OLS
6 First, I run the IV regression and obtain a residual. Second, I regresses the residual on all exogenous variables and obtains the R-squared. Third, the number of observations times the R-squared is distributed 2 χq , where q is the number of IV variables outside of the model minus the number of endogenous variables. If we reject the null hypotheses, under the assumption that all instrumental variables are not correlated with the error term in the structural model, we know that some of the instruments are not exogenous. See Wooldridge (2006, p.533-34) for testing over identifying restrictions. 110 estimates provide a lower bound and IV estimates provide an upper bound for the actual value of the estimated coefficient (Markussen, 2008).
3.5 Conclusion
There are only a few microeconomic studies that focus on how de jure land rights affect household well-being. This Chapter studies how de jure land rights affect household consumption expenditures. Although household consumption expenditure is not a perfect measure of household well-being, it has been widely used to measure household welfare in developing countries due to problems with other measures of well- being, such as annual income.
Both OLS and IV estimates are used to explore the link between de jure land rights and household consumption. IV estimation is used to account for selection effects in household participation in Cambodia’s programs to convert de facto land claims into formal de jure land rights supported by a national land registry. Three measures of land rights—titles, titles and application receipts, and application receipts—are used in the analysis to explore potentially different impacts of each type of land right on household expenditure. OLS estimates indicate a positive and statistically significant link between land rights and household consumption. The effect becomes substantially larger when instruments for land rights are used. The larger impact is expected when the endogenous variable is binary and instruments are correlated with measurement error of the dependent variable. In this case, OLS estimates provide a lower bound and IV estimates provide an upper bound for the actual value of the estimated coefficient.
111
The magnitudes of the effect of land rights on household consumption vary with different measures of land rights. All estimated coefficients on the three measures of land of land rights are positive and statistically significant. In sum, land titles have increased household welfare in Cambodia.
112
Table 3.1
Village characteristics: Summary statistics
Mean Std. Dev. Village characteristics Total land area (hectares) 765.45 1677.91 Total agricultural land area (hectares) 222.37 352.14 Irrigated land area (hectares) 37.83 103.97
Distance from village to All -weather road (Km) 2.64 13.51 Bus stop (Km) 17.30 33.23 Taxi stop (Km) 5.42 15.03 Village head office (Km) 10.49 16.57 District head office (Km) 29.68 29.86
Households with Electricity (%) 0.27 0.39 Clean water (%) 0.18 0.35
Amount of rain and crop damage Normal rain fall (%) 0.27 0.44 Crop damage (%) 0.19 0.40
Development projects Conducted by the government (%) 0.40 0.50 Number of observations 900 Note: The composition of development projects is as follows: 50% out of the 40% reported above is agriculture, 30% infrastructure, 12% education, 4% health, and 3% water related. Source: All data are from CSES (2004).
113
Table 3.2
Household loans: Summary statistics
Mean Std. Dev. Purpose of loan: Agricultural purpose (%) 0.24 0.43 Household consumption (%) 0.32 0.47 Household medical expense (%) 0.12 0.33
Loans obtained from: Relative in Cambodia (%) 0.27 0.44 Friend (%) 0.14 0.35 Money lender (%) 0.25 0.43 Bank (%) 0.036 0.18 NGOa (%) 0.16 0.37 Amount borrowedb (Reil) 953,324 3,706,033 Monthly interest rate (%) 0.04 0.08 Number of observations 8,024 Note: a. NGO stands for non-governmental organization. b. The Reil is the Cambodian currency. Source: All data are from CSES (2004).
114
Table 3.3
Household Characteristics: Summary statistics
Mean Std. Dev. Head of household characteristics Education of household head (years) 4.37 4.02 Age of head (years) 45.14 13.74 Female headed (%) 0.22 0.41 Household size (number) 4.94 1.98 Number of males age 10 to 14 years old 0.36 0.60 Number of females age 10 to 14 years old 0.35 0.59 Number of males age 15 to 55 years old 1.29 0.94 Number of females age 15 to 55 years old 1.45 0.90 Number of males age over 55 years old 0.19 0.39 Number of females age over 55 years old 0.26 0.45 Number of household members away from home a 0.09 0.40 Head is self employed (%) 0.71 0.45 Head is paid employee (%) 0.19 0.39 Did not have enough food to consume in 2004 (%) 0.16 0.36 Total household net income per month (Reil)b 372,492.2 3,339,819
Health status of head Good health (%) 0.15 0.35 Average health (%) 0.67 0.47 Bad health (%) 0.18 0.38
Number and size of farmlands Number of farmlands 1.31 1.24 Size of operated farmlands 1.29 0.45 Landless households (%) 0. 29 0.45
Household consumption expenditures Total consumption expenditure per month (Reil) 516,800 661,520 Total food consumption expenditure per month (Reil) 271,644 204,560 Number of observations 14,984 Note: a. Members who were away from home for more than two months in the past five years. b. Cambodian currency. Source: All data are from CSES (2004).
115
Table 3.4
Land titles and household consumption: OLS and instrumental variable regressions
Log of total monthly Log of total monthly
expenditure (OLS) expenditure (IV) Households with land title (%) 0.024 (1.90) 0.465 (4.74)
Education (years) 0.030 (11.47) 0.029 (10.87)
Education squared (*1000) -0.068 (-3.29) -0.68 (-3.34)
Age (years) 0.009 (4.47) 0.008 (3.34)
Age squared (* 1000) -0.056 (-3.18) -0.007 (-2.45)
Female head (%) -0.147 (-13.69) -0.144 (-12.95)
Self employed (%) -0.018 (-1.10) -0.025 (-1.48)
Paid employee (%) -0.060 (-2.98) -0.053 (-2.53)
Number of males age 0 to 4 0.061 (8.51) 0.059 (7.90)
Number of females age 0 to 4 0.063 (9.22) 0.064 (8.93)
Number of males age 10 to 14 0.058 (8.92) 0.056 (8.30)
Number of females age 10 to 14 0.067 (9.98) 0.063 (8.95)
Number of males age 15 to 54 0.125 (26.23) 0.120 (25.53)
Number of females age 15 to 54 0.118 (25.56) 0.117 (24.18) Number of members away from home for more -0.048 (-4.58) -0.048 (-4.68) than two months Total value of durable goods (*1000) 0.020 (12.80) 0.020 (12.96)
Size of operated farm land (hectares) 0.012 (4.25) 0.007 (3.09)
Size of operated farm land squared (*1000) -0.025 (-4.42) -0.019 (-3.41)
Access to all weather roads (km) -0.002 (-3.27) -0.002 (-3.19)
Type of land dummies Yes Yes
Village dummies Yes Yes
Over identification test (p-value) 0.97
R2 0.65 0.62
Observations 14,976 14,976 Note: All expenditures are in logs. t-statistics are in parenthesis. The regression is reported with robust standard errors. Source: All data are from CSES (2004).
116
Table 3.5
A combined land rights and household consumption: OLS and instrumental variable regressions
Log of total monthly Log of total monthly
expenditure (OLS) expenditure (IV) Households with land title or application receipt (%) 0.046 (4.19) 0.192 (3.35)
Education (years) 0.030 (11.33) 0.028 (10.50)
Education squared (*1000) -0.60 (-3.21) -0.60 (-3.03)
Age (years) 0.009 (4.44) 0.007 (3.83)
Age squared (*1000) -0.060 (-3.03) -0.059 (-2.58)
Female head (%) -0.146 (-13.62) -0.143 (-13.28)
Self employed (%) -0.021 (-1.27) -0.028 (-1.70)
Paid employee (%) -0.057 (-3.04) -0.050 (-2.69)
Number of male aged 0 to 4 0.061 (8.51) 0.060 (8.19)
Number of female aged 0 to 4 0.064 (9.24) 0.064 (8.73)
Number of male aged 10 to 14 0.058 (8.92) 0.057 (8.19)
Number of female aged 10 to 14 0.066 (9.81) 0.063 (8.91)
Number of male aged 15 to 54 0.123 (26.16) 0.121 (26.74)
Number of female aged 15 to 54 0.118 (25.41) 0.116 (25.32) Number of members away from home for more than -0.047 (-4.357) -0.046 (-4.25) two months Total value of durable goods (*1000) 0.002 (12.82) 0.002 (12.88)
Size of operated land (hectares) 0.011 (4.04) 0.008 (3.28)
Size of operated farm land squared (*1000) -0.002 (-4.53) -0.002 (-3.38)
Access to all weather roads (km) -0.002 (-3.44) -0.002 (-3.79)
Type of land dummies Yes Yes
Village dummies Yes Yes
Over identification test (p-value) 0.94
R2 0.65 0.65
Observations 14,976 14,976 Note: All expenditures are in logs. t-statistics are in parenthesis. The regression is reported with robust standard errors. Source: All data are from CSES (2004). 117
Table 3.6
Application receipts and household consumption: OLS and instrumental variable regressions
Log of total monthly Log of total monthly
expenditure (OLS) expenditure (IV) Households with an application receipt (%) 0.049 (3.54) 0.25 (3.37)
Education (years) 0.030 (10.37) 0.029 (9.99)
Education squared (time 1000) -0.60 (-2.70) -0.54 (-2.51)
Age (years) 0.003 (3.62) 0.007 (3.12)
Age squared (time 1000) -0.050 (-2.31) -0.048 (-2.00)
Female head (%) -0.136 (-11.51) -0.133 (-11.30)
Self employed (%) -0.018 (-1.01) -0.024 (-1.38)
Paid employee (%) -0.059 (-2.89) -0.053 (-2.64)
Number of males age 0 to 4 0.065 (8.31) 0.065 (8.23)
Number of females age 0 to 4 0.058 (7.53) 0.058 (7.59)
Number of males age 10 to 14 0.054 (7.51) 0.054 (7.48)
Number of females age 10 to 14 0.066 (8.66) 0.063 (8.24)
Number of males age 15 to 54 0.122 (23.90) 0.121 (23.63)
Number of females age 15 to 54 0.118 (23.90) 0.116 (22.47) Number of members away from home for more -0.048 (-4.10) -0.047 (-4.10) than two months Total value of durable goods (*1000) 0.019 (12.24) 0.019 (12.54)
Size of operated land (hectares) 0.026 (6.51) 0.021 (5.52)
Size of operated farm land squared (*1000) -0.016 (-3.95) -0.016 (-3.82)
Access to all weather roads (km) 0.002 (1.11) 0.001 (0.89)
Type of land dummies Yes Yes
Village dummies Yes Yes
Over identification test (p-value) 0.95
R2 0.67 0.66
Observations 12,530 12,530 Note: All expenditures are in logs. t-statistics are in parenthesis. The regression is reported with robust standard errors. Source: All data are from CSES (2004). 118
Table 3.7
Regression of land rights on all instruments and exogenous variables
Land title or Receipt Land title receipt Given by state (%) 0.089 (4.53) 0.173 (13.58) 0.164 (11.91) Inherited from families (%) 0.052 (4.74) 0.086 (7.38) 0.077 (6.15) Bought land (%) 0.173 (15.26) 0.237 (20.03) 0.168 (12.14) Occupied for free (%) -0.040 (-3.25) -0.112 (-7.82) -0.081 (-5.69) Received from friend (%) -0.116 (-8.56) -0.305 (-19.54) -0.206 (-13.12) Number of year owned land 0.002 (4.53) 0.003 (5.56) 0.002 (3.29) Education (years) -0.000 (-0.44) 0.002 (1.83) 0.003 (2.11) Education squared 0.000 (0.70) -0.000 (-1.73) -0.000 (-2.15) Age (years) 0.000 (0.53) 0.002 (1.88) 0.002 (1.57) Age squared -0.000 (-0.31) -0.000 (-1.43) -0.000 (-1.22) Female head (%) 0.006 (1.02) 0.007 (0.98) 0.007 (1.00) Self employed (%) -0.014 (-1.52) -0.017 (-1.77) -0.009 (-0.95) Paid employee (%) 0.007 (0.77) -0.006 (-0.61) -0.007 (-0.67) Number of males age 0 to 4 0.005 (1.01) 0.004 (0.84) 0.000 (0.15) Number of females age 0 to 4 -0.003 (-0.67) -0.003 (-0.86) -0.003 (-0.59) Number of males age 10 to 14 0.001 (0.32) -0.000 (-0.02) -0.001 (-0.23) Number of females age 10 to 14 0.003 (0.88) 0.012 (2.56) 0.011 (2.35) Number of males age 15 to 54 0.003 (1.17) 0.002 (0.59) -0.001 (-0.42) Number of females age 15 to 54 -0.001 (-0.50) 0.002 (0.63) 0.002 (0.88) Number of members away from home for 0.000 (0.01) -0.006 (-0.82) -0.002 (-0.37) more than two months Total value of durable goods 0.000 (1.69) 0.000 (3.09) 0.000 (2.47) Size of operated farm land (hectares) 0.002 (2.16) 0.004 (3.76) 0.002 (1.09) Size of operated farm land squared -0.000 (-1.14) -0.000 (-3.52) -0.000 (-0.74) Access to all weather roads (km) -0.000 (-1.70) -0.000 (-0.02) 0.001 (0.00) Type of land dummies Yes Yes Yes Village dummies Yes Yes Yes F-test on significance of instruments (p-value) 0.00 0.00 0.00 R2 0.43 0.57 0.54 Observations 14,973 14,973 12,527 Note: The regression is reported with robust standard errors. Source: All data are from CSES (2004).
119
References
Alston, L., Libecap, G., & Mueller, B. (1999). Titles, conflict, and land use. Ann Arbor: University of Michigan Press.
Alchian, A., & Demsetz, H. (1973). The property rights paradigm. Journal of Economic History, 33(1), 16-27.
Angrist, J. D., & Krueger, A. B. (2001). Instrumental variables and the search for identification: From supply and demand to natural experiments. Journal of Economic Perspectives, 15(4), 69-85.
Barraclough, S. (1991). An end to hunger? The Social origins to food strategies. Geneva: Institute for Social Development.
Besley, T. (1995). Property rights and investment incentives: Theory and evidence from Ghana. Journal of Political Economy, 10(5), 903-937.
Besley, T., & Burgess, R. (2000). Land reform poverty and growth: Evidence from India. Quarterly Journal of Economic, 115(3), 389-430.
Brasselle, A., Gaspart, F., & Platteau, J. P. (2002). Land tenure security and investment incentive puzzling evidence from Burkina Faso. Journal of Development Economics, 67(2), 373-418.
Carter, M., & Olinto, P. (2003). Getting institutions “right” for whom? Credit constraints and the impact of property rights on the quantity and composition of investment. American Journal of Agricultural Economics, 85(1), 173-186.
Chan, S., Tep, S., & Sarthi, A. (2001). Land tenure in Cambodia: A data update (CDRI Working Paper 19). Phnom Penh: CDRI.
Cooper, G. (2002). Land policy and conflict in Cambodia. Unpublished manuscript.
Damme, W. V., Leemput, L. V., Por, I., Hardeman, W., & Meessen, B. (2004). Out-of- pocket health expenditure and debt in poor households: Evidence from Cambodia. Tropical Medicine and International Health, 9(2), 273-280.
Deaton, A. (1997). The analysis of household survey. Baltimore, Maryland: The Johns Hopkins University Press.
Deaton, A., & Zaidi, S. (2002). A guide to aggregating consumption expenditures, living standards measurement study (Working Paper 135). Washington, D.C: The World Bank.
120
Deininger, K. (2003). Land policy for growth and poverty reduction. Washington, D.C: The International Bank for Reconstruction and Development/ the World Bank.
Deininger, K., & Jin, S. (2006). Tenure security and land-related investment: Evidence from Ethiopia. European Economic Review, 50(5), 1245-1277.
Demsetz, H. (1967). Toward a theory of property rights. American Economics Review, 57(1), 347-359.
De Sotto, H. (2000). The mystery of capital: Why capitalism triumphs in the West and fails everywhere else. New York: Basic Books.
Feder, G., & Nishio, A. (1998). The benefits of land registration and titling: Economic and social perspectives. Land Use Policy, 15(1), 25-43.
Feder, G., Onchan, T., Chanlamwong, Y., & Hongladarom, C. (1988). Land policies and farm productivity in Thailand. Baltimore: Johns Hopkins University Press.
Fenske, J. (2011). Land tenure and investment incentives: Evidence from West Africa. Journal of Development Economics, 95(2), 137-156.
Field, E. (2007). Entitled to work: Urban property rights and labor supply in Peru. The Quarterly Journal of Economics, 122(4), 1562-1602.
Field, E., & Torero, M. (2006). Do property titles increase credit access among the urban poor? Evidence from a nationwide titling program. Mimeo, Harvard University.
Finan, F., Sadoulet, E., & de Janvry, A. (2005). Measuring the poverty reduction potential of land in rural Mexico. Journal of Development Economics, 77(1), 27- 51.
Galiani, S., & Schargrodsky, E. (2004). Effects of titling on child health. Economic of Human Biology, 2(3), 353-372.
Galiani, S., & Schargrodsky, E. (2010). Property rights for the poor: The effects of land titling, Journal of Public Economics, 94(9-10), 700-729.
Garderen, K. J. V., & Shah, C. (2002). Exact interpretation of dummy variables in semi- logarithmic equation. Journal of Econometrics, 5(1), 149-159.
Gibson, J. (2002). Why does the Engel method work? Food demand, economies of size and household survey methods. Oxford Bulletin of Economics and Statistics, 64(4), 341-59.
Goldstein, M., & Udry, C. (2008). The profits of power: Land rights and agricultural investment in Ghana. Journal of Political Economy, 116(6), 981-1022.
121
Huy, R., Wichmann, O., Beatty, M., Ngan, C., Duong, S., Margolis, H. S., & Vong, S. (2009). Cost of dengue and other febrile illnesses to households in rural Cambodia: A prospective community-based case-control study. BMC Public Health, 9, 1-55.
Jacoby, H., Li, G., & Rozelle, S. (2002). Hazards of expropriation: Tenure insecurity and investment in rural China. The American Economic Review, 92(5), 1420-1447.
Jacoby, H., & Minten, B. (2007). Land titles, investment, and agricultural productivity in Madagascar. The World Bank Economic Review, 21(3), 461-485.
Kane, T. J., Rouse, C. E., & Staiger, D. (1999). Estimating returns to schooling, when schooling is misreported (NBER Working Paper no. 7235).
Kenjiro, Y. (2005). Why illness causes more serious economic damage than crop failure in rural Cambodia. Development and Change, 36 (4), 759-783.
Ledgerwood, J. (1998). Rural development in Cambodia: The view from the village. In F. G. Brown, & D. G Timberman (Eds), Cambodia and the international community: The quest for peace, development, and democracy (pp. 127-147). New York: Asia Society.
Lopez, R., & Valdes, A. (2000). Fighting rural poverty in Latin America: New evidence of the effects of education, demographics, and access to land. Economic Development and Cultural Change, 49 (1), 197-211.
Maxwell, D., & Wiebe, K. (1993). Land tenure and food security: Exploring dynamic linkages. Development and Change, 101(2), 223-224.
Markussen, T. (2008). Property rights, productivity, and common property resources: Insights from Rural Cambodia. World Development, 36(11), 2277-2296.
Migot-Adholla, S., Hazell, P., Blarel, B., & Place, F. (1991). Indigenous land rights systems in Sub-Saharan African: A constraint on productivity. The World Bank Economic Review, 5(1), 155-175.
National Institute of Statistics. (2004). CSES 2003-2005 data management manual. Phnom Penh: Cambodia
Place, F., & Hazell, P. (1993). Productivity effects of indigenous land tenure systems in Sub-Saharan African. American Journal of Agricultural Economics, 75(1), 10-19.
122
Place, F., & Migot-Adholla, S. (1998). The economic effects of land registration on small holder farms in Kenya: Evidence from Nyeri and Kakamega districts. Land Economics, 74(3), 360-373.
Ravallion, M. (1996). Issues in measuring and modeling poverty. Economic Journal, 106 (438), 1328-1343.
Royal Government of Cambodia (2009). National strategic development plan update 2009-2013. Phnom Penh: Cambodia.
Sargan, J. D. (1958). The estimation of economic relationships using instrumental variables. Econometrica, XXVI, 393-415.
So, S. B. (2009). Political economy of land registration in Cambodia. (Doctoral dissertation). Retrieved from ProQuest Dissertations and Theses. (Accession Order No. AAT 3359039)
Wooldridge, J. (2002). Econometric analysis of cross section and panel data. Cambridge, Massachusetts: The MIT Press.
Wooldridge, J. (2006). Introductory econometrics a modern approach. Ohio: Thomson South-Western.
World Bank. (1975). Land reform: Sector policy paper. Washington, D.C: World Bank.
______. (2002). LMAP project appraisal documents report no. 22869-KH. Phnom Penh: World Bank.
______. (2006). Cambodia halving poverty by 2015? Poverty assessment 2006. Phnom Penh: World Bank.
______. (2008). Cambodia at a Glance, 24 September 2008. Retrieved February 15, 2009, from http://devdata.worldbank.org/AAG/khm_aag.pdf .
123
CHAPTER FOUR
THE IMPACT OF PROPERTY RIGHTS IN LAND ON CHILD HEALTH
4.1 Introduction
This chapter explores the relationship between de jure land rights and child health
in Cambodian urban and rural households with land holdings. The relationship between
de jure land rights and children’s health can result from several distinct types of benefits
of land registration. Previous studies have documented a strong correlation between
socio-economic status and child health (Currie & Stabile, 2003). Theoretically,
establishing systems of formal land rights and the resulting improvement in tenure
security can lead to a range of social benefits including increased investment in land
(Abdulai, Owusu, & Goetz, 2011; Deininger & Jin, 2006), improved access to credit
markets (Feder, Onchan, Chanlamwong, & Hongladarom, 1988; Field & Torero, 2006),
increases in the participation of adult labor and decreases in the participation of child
labor (Field, 2007), and increases in human capital, housing improvements and
maintenance, and children’s health in urban squat households (Galiani & Schargrodsky,
2004; Galiani & Schargrodsky, 2010).
There are currently few studies that focus on how the establishment of land rights
affects children’s health. Galiani and Schargrodsky (2003, 2004) argued that land titles
could have positive indirect impacts on children’s health because they lead to more
investment in housing and household structures. They posit that “under the security provided by proper land titles, families are more likely to invest in improvements in
water distribution within their homes, treatment of fecal evacuation, treatment of garbage
124
disposal, safety and quality of heating systems, air ventilation conditions, and quality of
the soil. Moreover, the possibility of exchanging houses when their size is inadequate for
family needs may help to avoid overcrowding” (p. 354). These investments have been shown to contribute to increased child health. Using data from a 1984 land reform in a poor Buenos Aires suburb in which de jure rights were exogenously assigned to households in their sample, they found “that in the titled parcels children enjoy better weight-for-height z-scores (but similar height-for-age z-scores), and teenage girls have lower pregnancy rates than those in untitled parcels” (p. 353). A similar study for Peru highlighted the same results but also found that titles were associated with an increase in the risk that the household’s children would be overweight or obese (Vogl, 2007).
This study examines whether the establishment of de jure land rights in Cambodia caused changes in child health in Cambodian urban and rural households with land holdings. The econometric model builds on a second study, Galiani and Schargrodsky
(2004), who directly estimated the impact of establishing de jure land rights on child health, as measured by weight-for-height and height-for-age variables. This study estimates this relationship for a sample of households drawn from the 2003-2004
Cambodia Socio-Economic Survey (CSES) that includes information on 8,745 children below six years of age. The study uses three alternate measures of de jure land rights collected by CSES for each household in the sample. De jure rights to a household’s land claims were not exogenously imposed but were awarded when and if a household completed a costly and complex bureaucratic process to establish title in the land. My analysis uses instrumental variable regression techniques to account for the selection
125
effects in a household’s decision to participate in a land titling program to convert de
facto land claims into formal de jure land rights.
The analysis in this study contributes to this field’s empirical literature by broadening the scope of study to include the effects of land titling programs on child health in both urban and rural areas. Galiani and Schargrodsky (2004) and Vogl (2007)
estimated the effect of residential land title on child health in households living in urban
squats whereas this study examines the effect of de jure rights on child health in
households living in both rural and urban areas. In this study results from OLS
regressions indicate that children living in households with de jure land rights have similar weight-for-height and height-for-age z-scores as children living in households with de facto land rights. Use of instrumental variable regression techniques to account for endogeneity of land rights yields essentially the same outcomes weight-for-height regressions but different outcomes height-for-age regressions: Children living in households with titled parcels have better height-for-age z-scores than children living in households with untitled parcels. The study’s results may contribute to illuminating policy questions regarding the ongoing implementation of land titling programs in
Cambodia.
4.2 The impact of land property rights on children’s health
The effect of de jure land rights on various social indicators such as land investment, crop productivity, access to credit markets, and land values have been widely studied in the past two decades, but only recently have a few studies focused on how de jure land titles affect child health. Galiani and Schargrodsky (2003, 2004) showed that
126 land titles in Argentina’s urban squats could have positive indirect impacts on child health because they lead to more investment in housing and household structures. They asserted that “under the security provided by proper land titles, families are more likely to invest in improvements in water distribution within their homes, treatment of fecal evacuation, treatment of garbage disposal, safety and quality of heating, air ventilation conditions, and quality of the soil. Moreover, the possibility of exchanging houses when their size is inadequate for family needs may help to avoid overcrowding” (p. 354).
Their study indicated that in the titled parcels children have higher weight-for- height z-scores--an indicator of current nutrition status--but similar height-for-age z- scores, an indicator of cumulative growth deficits. Further, teenage girls living in households with titled land parcels had a lower pregnancy rate than girls living in households with untitled land parcels. Another study on child health and nutritional status in urban areas of Peru, using data from a survey of 27 Lima communities conducted in
2004, showed that land titles improved children’s weights but not their heights (Vogl,
2007). Vogl asserted that the positive effects of land rights in urban areas of Peru on child nutrition might be due to increases in labor force participation, housing investment, and credit access. A previous study of an urban land titling programs in Peru (Field 2007) showed that adults living in households with titled parcels increase their labor force participation because they do not have to keep people at home to guard their informal properties. Field found that weekly household labor hours on titled parcels were 16-17 hours higher than on untitled parcels. Vogl (2007) maintained that increases in income from an additional member participating in the labor market might improve child nutrition but that working could impact time spent taking care of children if the labor
127
force participant is the child’s primary caregiver. Thus, the change in time allocation
might affect the child’s nutrition and care. “The direction of this change depends on the
balance of income and substitution effects” (Volg, 2007, p. 303). Further, Field (2005)
and Field and Torero (2006) showed that housing investment and credit access increased
among those households with titled parcels. Empirical findings from Volg (2007) and
Galiani and Schargrodsky (2004) are similar: Children living in households with titled
parcels have better weight-for-height z-scores but the same height-for-age z-scores.
However, Volg (2007) highlighted that de jure title increased the risk to a household’s
children of being overweight or obese.1 This study only examines the effect of land rights
on child nutritional status in Cambodia measured by weigh-for-height z-scores and
height-for-age z-scores, as there were few overweight children in Cambodia in 2003-
2004. The survey data from the 2003-2004 CSES shows that approximately 0.2 percent
of children had a Body Mass Index (BMI) more than 25.2 O’Donnell, Doorslaer,
Wagstaff, and Lindelow (2008) maintained that the incidence of overweight or obesity is
relatively low in South and East Asia.
De jure land rights in Cambodia may have impact on children’s health through increases in household consumption, land investment, and housing investment. The empirical results from Chapters Two and Three indicate that de jure land rights enhance land investment and raise household consumption. Secure land rights facilitate smoothing household consumption, thereby avoiding low consumption periods that may also be
1 Vogl argued that the observed weight gain does not imply an improvement in the child’s nutritional status (p. 302). A study of gender and land ownership in Nepal showed that children of mothers who own land are less likely to be severely underweight (Allendorf, 2007). 2 I use the adult BMI cutoff for children. A person is considered to be overweight if their BMI exceeds 25, normal if their BMI is between 18.5 to 24.9, and underweight if their BMI is below 18.49 (O’Donnell et al., 2008). 128
associated with lower nutrient intakes for the children in the household. Further, the household’s additional income generated from its higher land investment may also improve children’s nutrition. On the other hand, a household with de jure titles may have the ability to exchange its house when its size is inadequate for family needs or to use their increased access to credit markets to finance improvements in their housing structures. The investment in housing or improvement in household structures as articulated in Galiani and Schargrodsky (2004) may have an indirect impact on child health.
4.3 Children’s health in Cambodia
In 2003-2004, Cambodia was the poorest country in Asia despite considerable
growth over the previous decade (CDHS, 2005). Cambodia’s population had the worst
health in the Western Pacific Region and its overall health system performance was
classified 174th among members of the World Health Organization (CDHS 2000; WHO,
2000). Life expectancy at birth for males was 58 years and for females was 64 years
(CDHS 2005). One in ten children died before their fifth birthday due to insufficient
health care services, poor hygiene, and poverty (CDHS 2000). Nearly half of Cambodian
children were malnourished in their second year of life (HKIC, 2002). Preventable
diseases, such as diarrhea, acute respiratory infections, dengue hemorrhagic fever,
malaria, and malnutrition have affected the lives of many children in Cambodia (CDHS,
2005).
Although Cambodia is a poor country with a fragile health system, the 2005
Health Demographic Survey (CDHS) showed that health outcomes of children had
129
improved. Infant mortality declined from 95 in 2000 to 66 in 2005 per 1,000 live births.
The mortality rate of under-five children decreased from 124 to 83 over the same period.
The percentage of children under five years of age who were stunted declined from 45
percent in 2000 to 37 percent in 2005, while the percentage of children who were wasted
declined from 15 percent in 2000 to 7 percent in 2005, and the percentage underweight
declined from 45 percent in 2000 to 36 percent in 2005.3 The rate of primary school
attendance rose from 68 percent in 2000 to 77 percent in 2005.
Despite these improvements, inequality in health remains a critical issue for
Cambodia. Economic growth seems to benefit the rich more than the poor (Hong &
Mishra, 2006). About one third of the population is unable to pay medical bills from
either private or public providers. Annual per capita health expenditure is about US$37;
out-of-pocket expenditures are 68 percent of the total. Unexpected health expenditures
can easily drive people who live above the poverty line into poverty (MH, 2008). Several
qualitative studies indicate that expenditures for medical emergencies and household
consumption often are associated with a household’s sale of its lands and subsequently
becoming landless. Lands have often been sold at fire sale prices in case of emergency
needs (Kenjiro, 2005; Damme, Leemput, Por, Hardeman, & Meessen, 2004; Huy,
Wichmann, Beatty, Ngan, Duong, Margolis, & Vong, 2009).
3 The WHO found that amongst children who are under five years of age in Cambodia, 49.2 percent were stunted and 39.5 percent were underweight (WHO, 2007). The CHDS (2010) showed that 40 percent of children under five years of age were stunted and 11 percent were wasted. 130
4.4 Empirical analysis
4.4.1 Data
The 2003-2004 CSES incorporates a section on child health that allows an estimation of the effect of de jure land rights on some dimensions of child health. The nationwide survey provides information on 8,745 children less than 6 years of age, of which 3,900 are children under 2 years of age. Mothers or caregivers were asked about the child’s date of birth, immunization, nutrition intake, weight, height, vitamin A intake, and nighttime blindness. During the interview, the mother or caregiver was requested to answer questions related to immunization and nutrition intake for household children under two years of age. Interviewers first asked for an immunization card and recorded the dates and type of vaccinations: TB, polio, DTC/DPT and measles. If the card had been lost, the interviewer asked the mother or caretaker to recall types and approximate dates of vaccinations. Either the household head or spouse was asked about height, weight, and vitamin A intake of children less than 6 years of age. The interviewers also recorded the date of measurement and position of the child (standing up or lying down) while measuring their height. In the survey, 68 percent of the children were measured standing up, with the rest measured lying down. Height was recorded in centimeters and weight was recorded in kilograms.
The analysis in this study uses two variables—weight-for-height and height-for- age—to measure children’s health. These variables have been used in several empirical studies as measures of child health (Miller & Rodgers, 2009; Allendorf, 2007; Vogl,
2007; Galiani & Schargrodsky 2004; Hong & Mishra, 2004). Generally, both measures are used to assess nutritional status because they give “different information about growth
131
and body composition” (CDHS, 2000, p. 172). Weight-for-height measures body weight
in relation to body length or height. It reflects a short-term change in nutrition status and is generally used as a measure of current nutritional status. Children who have low
weight-for-height compared to children of the same age and sex in the reference population are considered thin and children who have very low weight-for-height are
considered wasted. The most commonly used cutoff to determine whether a child is malnourished or not is a z-score of minus two, i.e., two standard deviations below the
median of the reference population (O’Donnell et al., 2008). Children are too thin or
wasted if their weight-for-height measure is below minus two standard deviations and
severely wasted if their weight-for-height measure is below minus three standard
deviations. The height-for-age measure reflects a long-term inadequacy of nutrition.
Children who have low height-for-age compared to children of the same age and sex in the reference population are considered short and children who have very low height-for-
age are considered stunted. Children are stunted if their height-for-age measure is below minus two standard deviations and severely stunted if their height-for-age measure is below minus three standard deviations (O’Donnell et al., 2008; CDHS, 2000).
These anthropometric indicators are established by “comparing relevant measures with those of comparable individuals (in regard to age and sex) in the reference populations” (O’Donnell et al., 2008, p. 41). Before the release of a new international growth chart for children from birth to five years of age by the World Health
Organization (WHO) in 2006, the reference groups that were frequently utilized were based on a sample of the children in the United States. U.S reference groups were inappropriate for children in developing countries, who have different growth paths with
132
different feeding practices, led to a comprehensive review of the existing reference
groups and the development of a new reference group. The new international growth
chart developed by WHO in 2006 is considered more appropriate as it takes into account
the growth of children from different cultural settings (O’Donnell et al., 2008). Recently,
the Centers for Disease Control and Prevention (CDC) suggested that clinicians in the
United States use the 2006 WHO international growth charts for children ages less than
24 months and use the 2000 CDC growth charts for persons with ages 2-19 years
(Grummer-Strawn, Reinold, & Krebs, 2010). The 2006 WHO international growth charts
were based on data of healthy children from Brazil, Ghana, India, Norway, Oman, and
the United States (WHO, 2006). The 2000 CDC growth charts were created utilizing data
from five national surveys collected in the United States together with data extracted
from birth certificates and medical records (Vidmar, Carlin, Hesketh, & Cole, 2004).
The comparison of “relevant measures” with those of comparable individuals in the reference populations to create anthropometric indicators can be done in three ways: z-scores, percent of the median, and percentiles. A z-score is defined as “the difference between the value for an individual and the median value of the reference population for the same sex and age (or height), divided by the standard deviation of the reference population”. Z-scores are frequently used by researchers because they have desirable statistical properties. Z-scores can be used to describe means and standard deviation of population and sub-populations, a comparison that cannot be done with percentiles
(O’Donnell et al., 2008, p.42). “The percent of the median” is easier to compute than a z- score or percentile, while it neglects the distribution of the reference population around
133
the median (Gorstein, Sullivan, Yip, de Onis, Trowbridge, Fajans, & Clugston, 1994, p.
276).
In this analysis, the weight-for-height z-score and height-for-age z-score are computed using the 2006 WHO growth charts as the reference population. The 2006
WHO growth charts are more appropriate for this study as they have been calculated using children of diverse ethnic backgrounds and different cultural settings. Table 4.1 presents summary statistics for the weight-for-height and height-for-age z-scores for the sample of Cambodian children in the health section of the 2003-2004 CSES. The sample shows that 13 percent of children ages 0-6 have weight-for-height z-scores below minus two standard deviations and 41 percent have height-for-age z-scores below minus two standard deviations. This implies that 13 percent of children ages 0-6 are wasted and 41 are stunted. About 5 percent are severely wasted and 22 percent are severely stunted.
Figure 4.1 indicates the distributions of weight-for-height z-scores and height-for-age z- scores. Weight-for-height z-scores are skewed more to the right whereas height-for-age z- scores are concentrated more to the left, which show a prevalence of stunting.4
Table 4.2a reports summary statistics for children’s z-scores, children’s characteristics, and parental education when the data are sorted by the type of document under which the land parcel is held. Three categories of land documents are used in this study: Land title, application receipt, and no document. The weight-for-height z-scores for children living in households with titled parcels are similar to children living in households with untitled parcels but there are statistically significant differences between
4The percentage of children who were wasted and stunted in Cambodia declined in 2005 yet increased in 2010. The CDHS 2000 shows that 15 percent of children were wasted and 45 percent were stunted; the CDHS 2005 indicates 7 percent were wasted and 37 percent were stunted; the CDHS 2010 shows that 11 percent were wasted and 40 percent were stunted. 134 the groups when land titles and application receipts are combined in a single measure.
There are statistically significant differences in height-for-age z-scores for children living in households with titled and untitled parcels. The difference is also present when titles and application receipts are combined in a single measure.
Parental education is the same for land owners with title and de facto land owners.
Table 4.2b shows that educational attainments of the children’s parents are very low. On average, the child’s father has completed 4 years and the child’s mother 3 years of education.
This study uses three alternate measures of de jure land rights: land titles to their plot, measure combining titles and application receipts, and application receipts to their plot. As explained in Chapters Two and Three, land titles and application receipts may have different impacts on children’s health as the tenure security and other benefits provided by a land title and an application receipt are likely to differ. A de jure title can theoretically be used to secure loans from formal credit institutions that provide land owners with relatively low interest rates (compared to the informal loan market), whereas access to these institutions is impossible for a land owner with just an application receipt.
The application receipt can only be used to secure loans from a micro finance institution or a money lender in the informal credit market. With this limited access, farmers are less likely to be able to secure favorable loans for investment or to smooth household consumption. Further, without a land title, the market price of land is likely to be lower due to the higher risk of expropriation that potential buyers face. For these reasons, I conduct separate analyses of the impact of each land rights measure on children’s health.
135
4.4.2 Estimation strategy
Ruhm (2000) mentioned that child health hangs on several factors such as “the
stock of health capital, the level of medical technology, the price of and access to health
care, household income, and the time investments of parents” (p. 933-34). As discussed above, de jure land rights in Cambodia are most likely to affect child health through improvement in household income. This channel was carefully articulated by Volg
(2007). Higher income may improve household nutrition, provide the means for households to improve their housing structures, and increase their access to medical care.
Rosenzweig and Schultz (1983) posited that the empirical analysis of health
production models are mainly concentrated on estimating input demand functions or
reduced form health equations. These estimations are essential in providing policy-
relevant parameters and for prediction but the empirical analysis of the relationship
between health and health inputs have been hindered by the unavailability of data. In
return, “the equations (hybrids) with less desirable properties” have been estimated
(p.726).
The econometric model used to test the effect of de jure land rights on children’s
health in this study is based on Galiani’s and Schargrodsky’s (2004) model. Due to
unavailability of data on some inputs such as the price of medical care and access to
medical care, the child health specification here is a hybrid of the reduced form demand
function. The effect of land title on children’s health outcomes is specified as:
= 0 + 1 + 2 + 3 + (1)
𝑖𝑖ℎ 𝑖𝑖 ℎ ℎ 𝑖𝑖ℎ where 𝐻𝐻 𝛼𝛼measures𝛼𝛼 𝑋𝑋 the𝛼𝛼 health𝑍𝑍 𝛼𝛼 of𝑃𝑃 child𝜀𝜀 i in household h, is a vector of child i
𝑖𝑖ℎ 𝑖𝑖 characteristics𝐻𝐻 (gender, age, and relation to head), is a𝑋𝑋 vector of household h
𝑍𝑍ℎ 136
characteristics (father and mother education, father and mother employment), is a
ℎ binary variable denoting whether household h holds farm land with title, and 𝑃𝑃 is an
𝑖𝑖ℎ error term. 3 is the parameter of interest which highlights the impact of de jure𝜀𝜀 land rights on children’s𝛼𝛼 health of household h. The potential issue in estimating equation (1)
is the endogeneity of land rights. As discussed in Chapters Two and Three, families with
more education, resources, and power tend to have a higher probability of obtaining a
land title. Empirical studies of this issue in other developing countries have frequently
used an instrumental variable technique to account for selection effects with respect to
household participation in programs implemented to convert de facto land claims into
formal de jure rights. This study also uses an instrumental variable technique to account
for endogeneity of land rights.
4.4.3 Results
For each of the two health measures, equation (1) is estimated with and without
instrumentation for de jure land rights. The instrumental variables used in this Chapter’s
analysis are the same as those used in Chapter Three’s analysis: Mode of land acquisition
and length of land occupation. As discussed in more detail in Chapter Three, a valid
instrument must satisfy two conditions: The instrument must be correlated with the
measure of de jure land rights and be uncorrelated with the equation’s error term. The
survey indicates that some households received land from the government, purchased
land, inherited land from their families, received land from their friends or cleared empty
public land. The “mode of land acquisition” variables indicate the method by which the
household obtained their land plot. I argue that different types of mode of acquisition
137
were associated with different degrees of land tenure security. For example, households
who obtained land from the government or purchased it have a higher chance of
obtaining de jure title. As mentioned in Chapter One, starting from 1989, land in
Cambodia was privatized. The government allocated the solidarity group land to
households according to family size and availability of land in each village. The land
titling program was part of the 1989 land reform in which de facto land owners were
eligible to apply for de jure titles for their residential land and possession rights for their
cultivated land. Households who obtained land through this allocation might receive an
official document such as an application receipt that could serve as a proof of ownership
in a future dispute. Further, households who purchased land from their neighbors or
outsiders also tended to have some type of document such as a sales agreement between
buyer and seller witnessed at the village or commune offices as a proof of land
transaction. In general a buyer only wants to purchase land with secure rights. Therefore,
a seller has to show some type of document as proof of land ownership. On the contrary,
households who acquired plots by clearing empty land or receiving it from friends were
less likely to apply for de jure rights, at least in part because of the more tenuous nature
of their de facto claims. Those households usually had no supporting documents for their
occupation. Thus, the probability of obtaining a de jure title was likely to be higher for those who obtained land from the government or purchased it than those who obtained land from their friends or cleared empty public land. Duration of tenure is also likely to affect whether a household acquires title. The length of time that a household has
occupied farm land affected its legal claim to their land. Basically, duration of tenure
increases the legitimacy of the household’s claim to its land. The 2001 land law allows
138
people who had already occupied land peacefully before the adoption of the law to
complete a five-year possession term and then to claim de jure ownership rights to their
land. Both instruments affect rights to a family’s plot of land but have no direct impact on
children’s health.
4.4.4 Weight-for-height
First, I consider the impact of de jure land rights on the weight-for-height z-score,
which is an indicator of current nutritional status. The regression specifications include
other factors influencing children’s health, such as children’s characteristics, parents’ characteristics, and village dummies. Village dummies are included to control for fixed differences across villages affecting the two measures of child health. Table 4.3a presents the outcomes from OLS estimations when land rights are measured by whether a land- holding household has a land title. The estimated coefficient on mother’s education is positive but is not statistically significant at the 10 percent level. This result is in line with several other studies that find that wasting is not substantially influenced by mother’s education as articulated in Miller and Rodgers (2009). Using the 2005 CDHS, Miller and
Rodgers (2009) indicated that the estimated coefficient on mother’s education is positive but statistically insignificant. Galiani and Schargrodsky (2004) found the same result.
Miller and Rodgers (2009) maintained that maternal characteristics have less influence on
wasting.
The estimated coefficient on father’s education follows the results for the
mother’s education, as it is statistically insignificant at the 10 percent level. Galiani and
Schargrodsky (2004) and Miller and Rodgers (2009) found similar results. The estimated
139
coefficient on parent’s occupation is not statistically significant at the 10 percent level.
The estimated coefficient on household wealth is positive and statistically significant at
the 5 percent level. Its small magnitude indicates that it has only a very small effect on
children’s current nutritional status.
The OLS regression results indicate that children living in households with titled
parcels have similar weight-for-height z-scores as children living in households with
untitled parcels. Table 4.3b presents estimation results when land rights are measured with a variable combining title and application receipt. The estimated coefficient on the combined measure of land rights remains statistically insignificant at the 10 percent level.
The result is the same when rights to the land are measured with an application receipt
(see Table 4.3c). Using instrumental variables for land rights, I find no significant differences in weight for height z-scores between children of owners with land title to their plot and without land title to their plots. The estimated coefficients on land rights are
not statistically significant at the 10 percent level for the three measures of land rights.
4.4.5 Height-for-age
Next, the estimation focuses on the effect of de jure land rights on the height-for-
age z-score, which provides a measure of cumulative nutritional inadequacies. Table 4.4a
provides results from OLS estimations when land rights are measured by land title.
Again, the estimated coefficient on mother’s education is positive and statistically
insignificant at the 10 percent level. When land rights are measured by application
receipt, the estimated coefficient on mother’s education is positive and statistically significant at the 10 percent level. Miller and Rodgers (2009) showed that the estimated
140
coefficient on mother’s education in Cambodia is positive and statistically significant
whereas outcomes from Galiani and Schargrodsky (2004) were mixed. In their study, the
estimated coefficient on mother’s education was not statistically significant in a full
sample of children but it was statistically significant at the 10 percent level in a sub-
sample of children ages 5-11. Miller and Rodgers (2009) also found that the estimated
coefficient on father’s education to be positive and statistically significant whereas this
study and Galiani and Schargrodsky (2004) did not.5
The results from OLS estimations indicate that children living in households with titled parcels have similar height-for-age z-scores as children living in households with untitled parcels. The results are the same for the rights variable combining titles and application receipts. When land rights are measured by application receipts, the estimated coefficient on land rights remains statistically insignificant.
The estimated coefficient on land rights becomes, however, statistically significant at the five percent level when land rights are instrumented with mode of acquisition and duration of tenure variables. When land rights are measured with a combined measure of title and receipt, the estimated coefficient on the land rights
variable remains statistically significant at the five percent level (Table 4.4b and 4.4c).
The outcomes from IV regression estimations show that children of land owners with a
land title are about 0.4 standard deviation taller than children of land owners without a
land title. The effect corresponds to an increase of 0.39 centimeter for a boy 48 months
old and 100 centimeter tall. I used Sargan tests to check whether the instrumental
variables are correlated with the error term in equation (1) for all three measures of land
5The regression estimation in Miller and Rodgers (2009) did not include a village fixed effect to account for differences in each village. 141
rights. Results are reported in Tables 4.3a, 4.3b, 4.3c, 4.4a, 4.4b, and 4.4c. P-values show that the instruments pass the over-identifying restriction test.
This study’s estimate of the effect of land rights on the two measures of child health finds results that are similar to and different from Galiani and Schargrodsky’s
(2004) study and Volg’s (2007) study. Their studies found that children of titled parcels have higher weight-for-height z-scores than children of untitled parcels. However, there are no statistically significant differences in height-for-age z-scores between the two groups. This study finds that children of titled parcels have higher height-for-age z-scores than children of untitled parcels but there are no significant differences in weight-for- height z-scores. The differences between the results in this study and the two studies above may be due to the length of time between the household surveys and when the household received its titles as well as the nature of the data. For Galiani and
Schargrodsky’s (2004) study, the survey was conducted in 2003 and they selected households who received land titled in 1989-1991 and in 1997-1998 for their study. By the time that the survey was conducted, some households had land titles for 13 years and some for 7 years. Whereas the survey in Volg’s (2007) study was conducted in the 2004, he selected a sample of households who received titles in 1996-1997 and in 2000-2003 for his study. Therefore, by the time of the survey, some households had land titles for 7-
8 years and some just 1-4 years.
This study’s survey was conducted in 2003-2004 and titling program was started in 1989. The 2003-2004 CSES did not include the date that the title or application was awarded to each household. Title or application could be awarded to a household anytime
142
between 1989 and 2003. Both surveys focused on urban land titling program whereas this
study’s survey covered both urban and rural land titling program.
4.5 Conclusion
The analysis in this Chapter contributes to the empirical study of the effect of de
jure land rights on children’s health, a topic that has not been widely studied. The
outcomes from OLS estimations indicate that children of titled parcels have similar
weight-for-height z-scores and height-for-age z-scores as children of untitled parcels. The
outcomes from IV estimations show that children living in households with titled parcels
have better height-for-age z-scores than children living in households with untitled
parcels; the differences in the weight-for-height z-scores of children livings in households
with titled and untitled parcels are, however, not statistically significant. Children of
titled parcels are 0.4 standard deviation taller than children of untitled parcel.
This study’s results are similar to but different from the results found in Galiani
and Schargrodsky’s (2004) study and Volg’s (2007) study. They found that land title has positive effects on children’s current nutritional status while this study found that land
title is associated with higher long-term nutritional status. The differences between the results in this study and the two studies above may be due to the length of time between
the household surveys and the receipt of title as well as differences in the nature of the
data. Their studies observe the effect of urban land titling programs on child health while this study broadens the scope by including the effects of land titling programs on child health in both urban and rural areas. Findings in this study suggest that a land title is associated with improvements in the long-term health of children in Cambodia.
143
Table 4.1
Descriptive statistics of z-scores for child nutritional status in Cambodia, 2003-2004
Weight-for-height z-scores Height-for-age z-scores Mean -0.49 -1.32 Standard deviation(S.D.) 1.74 2.14 Percent below < -2 S.D. 0.13 0.41 Percent below < -3 S.D. 0.05 0.22 Number of children 6,567 6,567 Source: All data are from CSES (2003-2004).
144
Figure 4.1
Distributions of weight-for-height and height-for-age z-scores in Cambodia, 2003-2004
WHZ .4 .3 .2 Density .1 0
-5 0 5 z-score
Density y
HAZ .4 .3 .2 Density .1 0
-5 0 5 z-score
Density y
Source: All data are from CSES (2003-2004).
145
Table 4.2a
Characteristics of children and their families by type of document
Titled Untitled Sig. Title Untitled Sig. Receipt Untitled Sig. plot plot or plot plot plot receipt plot Weight-for-height -0.48 -0.40 -0.48 -0.38 ** -0.48 -0.38 ***
Height-for-age -1.46 -1.32 *** -1.42 -1.31 *** -1.39 -1.31
Male children (%) 0.51 0.50 0.53 0.50 ** 0.54 0.50 **
Age of children (years) 3.07 2.94 * 3.03 2.93 ** 2.97 2.93
Relation to head
Son or daughter (%) 0.80 0.84 * 0.79 0.85 * 0.78 0.85 *
Grandson or granddaughter 0.17 0.14 ** 0.18 0.13 * 0.19 0.13 * (%) Education of father (years) 4.36 4.37 4.29 4.41 4.22 4.41
Education of mother (years) 3.11 3.04 3.06 3.05 3.01 3.05
Father is self employed (%) 0.70 0.63 * 0.71 0.61 * 0.72 0.61 *
Mother is self employed (%) 0.08 0.07 0.09 0.07 * 0.09 0.07 **
Father is paid employee (%) 0.13 0.20 * 0.12 0.22 * 0.10 0.22 *
Mother is paid employee 0.08 0.02 * 0.02 0.00 * 0.01 0.02 ** (%) Total value of durable goods 1.55 1.66 1.43 1.74 ** 1.32 1.74 ** (million Reil) Total value of live stock 1.97 1.28 * 1.91 1.15 * 1.85 1.15 * (million Reil) Number of rooms in the 1.31 1.30 1.29 1.31 1.26 1.31 *** house Access to all weather roads 2.76 3.87 *** 3.30 3.87 3.82 3.87 (km) Number of observations 8.744 7440
Note: a. “Statistically significant” refers to the differences in sample means for families with different types of documents for their plots. b. * significant at 1%; ** significant at 5%;*** significant at 10%. c. Reil is Cambodian currency. Source: All data are from CSES (2003-2004).
146
Table 4.2b
Characteristics of children and their families: Summary statistics
Means Std. Dev. Male children (%) 0.51 0.50 Age of children (years) 2.30 1.73 Relation to head Son or daughter (%) 0.83 0.38 Grandson or granddaughter (%) 0.16 0.36 Education of father (years) 4.35 3.96 Education of mother (years) 3.01 3.20 Father is self employed (%) 0.64 0.48 Mother is self employed (%) 0.08 0.27 Father is paid employee (%) 0.19 0.39 Mother is paid employee (%) 0.02 0.13 Number of observations 8,745 Source: All data are from CSES (2004).
147
Table 4.3a
Land title and children’s health: OLS and instrumental variable regressions
Weight-for- Weight-for-
height (OLS) height (IV) Land title (%) -0.043 (-0.65) 0.149 (0.47)
Male children (%) -0.069 (-1.77) -0.070 (-1.80)
Son or daughter of head (%) 0.072 (0.41) 0.085 (0.48)
Grandson or granddaughter of head (%) 0.141 (0.79) 0.147 (0.81)
Age of children (average) -0.174 (-13.51) 0.115 (6.34)
Education of father (years) -0.006 (-0.98) -0.008(-1.10)
Education of mother (years) 0.012 (1.52) 0.025(1.54)
Father is self employed 0.182 (1.87) 0.182 (1.87)
Mother is self employed 0.140 (1.22) 0.142 (1.24)
Father is paid employee 0.168 (1.59) 0.172 (1.63)
Mother is paid employee -0.087 (0.47) -0.072 (-0.40)
Total value of durable goods 0.000 (2.61) 0.000 (2.45)
Access to all weather roads (km) 0.002 (0.57) -0.005 (-1.00)
Over identification test (p-value) 0.72
Village dummies Yes Yes
R2 0.29 0.29
Number of observations 6,478 6,478 t-statistic in parentheses. Source: All data are from CSES (2003-2004).
148
Table 4.3b
Combined land rights measure and children’s health: OLS and instrumental variable regressions
Weight-for- Weight-for-
height (OLS) height (IV) Land title and application receipt (%) -0.043 (-0.65) 0.077 (0.48)
Male children (%) -0.069 (-1.77) -0.070 (-1.81)
Son or daughter of head (%) 0.074 (0.42) 0.084 (0.47)
Grandson or granddaughter of head (%) 0.132 (0.72) 0.146 (0.80)
Age of children (average) -0.174 (-13.49) -0.174 (13.51)
Education of father (years) -0.006 (-0.96) -0.006(-0.94)
Education of mother (years) 0.012 (1.60) 0.012 (1.57)
Father is self employed 0.182 (1.87) 0.181 (1.86)
Mother is self employed 0.140 (1.22) 0.142 (1.24)
Father is paid employee 0.168 (1.59) 0.175 (1.65)
Mother is paid employee -0.087 (-0.49) -0.074 (-0.41)
Total value of durable goods 0.000 (2.61) 0.000 (2.51)
Access to all weather roads (km) 0.003 (0.57) -0.012 (-2.44)
Over identification test (p-value) 0.73
Village dummies Yes Yes
R2 0.29 0.29
Number of observations 6,478 6,478 t-statistic in parentheses. Source: All data are from CSES (2003-2004).
149
Table 4.3c
Application receipt and children’s health: OLS and instrumental variable regressions
Weight-for- Weight-for-
height (OLS) height (IV) Application receipt (%) -0.008 (-1.11) 0.035 (0.14)
Male children (%) -0.077 (-1.77) -0.135 (-2.45)
Son or daughter of head (%) 0.075 (0.37) 0.081 (0.40)
Grandson or granddaughter of head (%) 0.129 (0.62) 0.136 (0.66)
Age of children (average) -0.194 (13.59) -0.193 (13.59)
Education of father (years) -0.007 (-1.01) -0.007(-0.95)
Education of mother (years) 0.008 (1.01) 0.009(1.07)
Father is self employed 0.211 (1.97) 0.211 (1.97)
Mother is self employed 0.129 (0.89) 0.232 (1.31)
Father is paid employee 0.127 (0.96) 0.184 (1.57)
Mother is paid employee 0.089 (0.40) 0.007 (0.04)
Total value of durable goods 0.000 (1.50) 0.000 (2.04)
Access to all weather roads (km) -0.005 (-0.90) -0.012 (-2.40)
Over identification test (p-value) 0.40
Village dummies Yes Yes
R2 0.32 0.32
Number of observations 5,481 5,481 t-statistic in parentheses. Source: All data are from CSES (2003-2004).
150
Table 4.4a
Land title and children’s health: OLS and instrumental variable regressions
Height-for-age Height-for-age
(OLS) (IV) Land title (%) -0.061 -(-0.73) 0.40 (1.79)
Male children (%) -0.113 (-2.29) -0.117 (-2.37)
Son or daughter of head (%) -0.454 (-2.04) -0.400 (-1.78)
Grandson or granddaughter of head (%) -0.259 (-1.13) -0.245 (-1.06)
Age of children (average) -0.570 (-35.01) -0.577 (-35.05)
Education of father (years) 0.006 (0.83) 0.006(0.81)
Education of mother (years) 0.015 (1.56) 0.013 (1.37)
Father is self employed 0.201 (1.63) 0.200 (1.62)
Mother is self employed 0.121 (0.83) 0.136 (0.94)
Father is paid employee 0.123 (0.92) 0.146 (1.09)
Mother is paid employee 0.086 (0.38) 0.149 (0.65)
Total value of durable goods 0.000 (1.50) 0.000 (1.10)
Access to all weather road (km) -0.000 (-0.86) 0.010 (1.56)
Over identification test (p-value) 0.62
Village dummies Yes Yes
R2 0.34 0.34
Number of observations 6,478 6,478 t-statistic in parentheses. Source: All data are from CSES (2003-2004).
151
Table 4.4b
Combined land rights measure and children’s health: OLS and instrumental variable regressions
Height-for-age Height-for-age
(OLS) (IV) Land title and application receipt (%) -0.114 (-1.68) 0.396 (2.35)
Male children (%) -0.111 (-2.26) -0.067 (-1.31)
Son or daughter of head (%) -0.460 (-2.07) -0.403 (-1.80)
Grandson or granddaughter of head (%) -0.267 (-1.16) -0.254 (-1.11)
Age of children (average) -0.570 (-34.99) -0.571 (-35.72)
Education of father (years) 0.007 (0.86) 0.007(0.86)
Education of mother (years) 0.015 (1.56) 0.015 (1.56)
Father is self employed 0.201 (1.64) 0.195 (1.59)
Mother is self employed 0.119 (0.82) 0.127 (0.88)
Father is paid employee 0.115 (0.86) 0.119 (0.90)
Mother is paid employee 0.077 (0.34) 0.079 (0.35)
Total value of durable goods 0.000 (1.54) 0.000 (1.22)
Access to all weather road (km) -0.001 (-0.90) 0.004 (1.05)
Over identification test (p-value) 0.68
Village dummies Yes Yes
R2 0.35 0.35
Number of observations 6,478 6,478 t-statistic in parentheses. Source: All data are from CSES (2003-2004).
152
Table 4.4c
Application receipt and children’s health: OLS and instrumental variable regressions
Height-for-age Height-for-age
(OLS) (IV) Application receipt (%) -0.192 (-2.11) 0.509 (1.86)
Male children (%) -0.127 (-2.32) -0.113 (-2.02)
Son or daughter of head (%) -0.664 (-2.22) -0.511 (-1.99)
Grandson or granddaughter of head (%) -0.349 (-1.32) -0.340 (-1.29)
Age of children (average) -0.566 (-31.29) -0.122 (-31.35)
Education of father (years) 0.006 (0.70) 0.004 (0.52)
Education of mother (years) 0.024 (2.25) 0.025 (2.30)
Father is self employed 0.219 (1.60) 0.216 (1.58)
Mother is self employed 0.115 (0.71) 0.143 (0.68)
Father is paid employee 0.055 (0.38) 0.115 (0.78)
Mother is paid employee 0.008 (0.03) 0.007 (0.29)
Total value of durable goods 0.000 (1.53) 0.000 (1.41)
Access to all weather roads (km) -0.001 (-0.28) 0.004 (0.64)
Over identification test (p-value) 0.25
Village dummies Yes Yes
R2 0.36 0.33
Number of observations 5,481 5,481 t-statistic in parentheses. Source: All data are from CSES (2003-2004).
153
References
Abdulai, A., Owusu, V., & Goetz, R. (2011). Land tenure differences and investment in land improvement measures: Theoretical and empirical analyses. Journal of Development Economics, 96(1), 66-78.
Allendorf, K. (2007). Do women’s land rights promote empowerment and child health in Nepal? World Development, 35(11), 1975-1988.
Currie, J., & Stabile, M. (2003). Socioeconomic status and child health: Why is the relationship stronger for older children?. American Economic Review, 93(5), 1813-1823.
Damme, W. V., Leemput, L. V., Por, I., Hardeman, W., Meessen, B. (2004). Out-of- pocket health expenditure and debt in poor households: evidence from Cambodia. Tropical Medicine and International Health, 9(2), 273-280.
Deininger, K., & Jin, S. (2006). Tenure security and land-related investment: Evidence from Ethiopia. European Economic Review, 50(4), 1245-1277.
Feder, G., Onchan, T., Chanlamwong, Y., & Hongladarom, C. (1988). Land policies and farm productivity in Thailand. Baltimore: Johns Hopkins University Press.
Field, E. (2007). Entitled to work: Urban property rights and labor supply in Peru. The Quarterly Journal of Economics, 122(4), 1562-1602.
Field, E. (2005). Property rights and investment in urban slums. Journal of the European Economic Association, 3 (2-3), 279-290.
Field, E., & Torero, M. (2006). Do property titles increase credit access among the urban poor? Evidence from a nationwide titling program. Mimeo, Harvard University.
Galiani, S., & Schargrodsky, E. (2003). Effects of land titling. Mimeo, Universidad Torcuato Di Tella.
Galiani, S., & Schargrodsky, E. (2004). Effects of titling on child health. Economics and Human Biology, 2(3), 353-372.
Galiani, S., & Schargrodsky, E. (2010). Property rights for the poor: The effects of land titling. Journal of Public Economics, 94(9-10), 700-729.
Gorstein, J., Sullivan, K., Yip, R., de Onis, M., Trowbridge, F., Fajans, P., & Clugston, G. (1994). Issues in the assessment of nutritional status using anthropometry. Bulletin of the World Health Organization, 72 (2), 273-283.
154
Grummer-Strawn, L. M., Reinold, C., & Krebs, N. (2010). Use of World Health Organization and CDC growth charts for children aged 0–59 months in the United States. Morbidity and Mortality Weekly Report, 59(PR-9), 1-13.
Helen Keller International/Cambodia. (2002). An overview of nutrition sector activities in Cambodia. Phnom Penh: Cambodia
Hong, R., & Mishra, V. (2006). The effect of wealth inequality on chronic under- nutrition in Cambodian children. Journal of Health, Population, and Nutrition, 24(1), 89-99.
Huy, R., Wichmann, O., Beatty, M., Ngan, C., Duong, S., Margolis, H. S., & Vong, S. (2009). Cost of dengue and other febrile illnesses to households in rural Cambodia: A prospective community-based case-control study. BMC Public Health, 9, 1-55.
Kenjiro, Y. (2005). Why illness causes more serious economic damage than crop failure in rural Cambodia. Development and Change, 36 (4), 759-783.
Miller, J., & Rodgers, Y. N. (2009). Mother’s education and children’s nutritional status: New evidence from Cambodia. Asian Development Review, 26(1), 131-165.
Ministry of Health. (2008). Health strategic plan 2008-2015. Phnom Penh, Cambodia.
National Institute of Statistics, Directorate General for Health (Cambodia), and ORC Macro. (2001). Cambodia Demographic and Health Survey 2010. Phnom Penh, Cambodia, and Calverton, Maryland USA: National Institute of Statistics, Directorate General for Health, and ORC Macro.
National Institute of Statistics, Directorate General for Health (Cambodia), and ORC Macro. (2001). Cambodia Demographic and Health Survey 2005. Phnom Penh, Cambodia, and Calverton, Maryland USA: National Institute of Statistics, Directorate General for Health, and ORC Macro.
National Institute of Statistics, Directorate General for Health (Cambodia), and ORC Macro. (2001). Cambodia Demographic and Health Survey 2000. Phnom Penh, Cambodia, and Calverton, Maryland USA: National Institute of Statistics, Directorate General for Health, and ORC Macro.
O’Donnell, O., Doorslaer, E. V., Wagstaff, A., & Lindelow, M. (2008). Analyzing health equity using household survey data. Washington, D.C: The international bank for reconstruction and development / The World Bank.
Rosenzweig, M. R., & Schultz, P. T. (1983). Estimating a household production function: Heterogeneity, the demand for health inputs, and their effects on birth weight. Journal of Political Economy, 91 (5), 723-746.
155
Ruhm, C. J. (2000). Parental leave and child health. Journal of Health Economics, 19(6), 931-960
Vidmar, S., Carlin, J., Hesketh, K., & Cole, T. (2004). Standardizing anthropometric measures in children and adolescents with new functions for Egen. The Stata Journal, 4(1), 50-55.
Vogl, T. (2007). Urban land rights and child nutritional status in Peru, 2004. Economics and Human Biology, 5(2), 302-321.
World Health Organization. (2007). World health statistics. Geneva: WHO.
World Health Organization. (2006). WHO child growth standards. Geneva: WHO.
World Health Organization (WHO). (2000). Health System: Improving Performance. Geneva: WHO.
156